Conference PaperPDF Available

Perturbance: Unifying Research on Emotion, Intrusive Mentation and Other Psychological Phenomena with AI


Abstract and Figures

Intrusive mentation, rumination, obsession, and worry, referred to by Watkins [1] as "repetitive thought" (RT), are of great interest to psychology. This is partly because every typical adult is subject to "RT". A critical feature of "RT" is of transdiagnostic significance—for example obsessive compulsive disorder, insomnia and addictions involve unconstructive “RT”. We argue that “RT” cannot be understood in isolation but must rather be considered within models of whole minds. Researchers must adopt the designer stance in the tradition of Artificial Intelligence augmented by systematic conceptual analysis [2]. This means developing, exploring and implementing cognitive-affective architectures. Empirical research on "RT" needs to be driven by such theories, and theorizing about “RT” needs to consider such data. We draw attention to H-CogAff theory of mind (motive processing, emotion, etc.) and a class of emotions it posits called perturbance (or tertiary emotions) [3,4], as a foundation for the research programme we advocate. Briefly, a perturbance is a mental state in which motivators tend to disrupt executive processes. We argue that grief, limerence (the attraction phase of romantic love) and a host of other psychological phenomena involving "RT" should be conceptualized in terms of perturbance and related design-based constructs. We call for new taxonomies of "RT" in terms of information processing architectures such as H-CogAff. We claim general theories of emotion also need to recognize perturbance and other architecture-based aspects of emotion. Meanwhile “cognitive” architectures need to consider requirements of autonomous agency, leading to cognitive-affective architectures.
Content may be subject to copyright.
Perturbance: Unifying research on emotion, intrusive
mentation and other psychological phenomena with AI
Luc P. Beaudoin1, Sylwia Hyniewska2 and Eva Hudlicka3
Abstract. Intrusive mentation, rumination, obsession, and
worry, referred to by Watkins [1] as "repetitive thought" (RT),
are of great interest to psychology. This is partly because every
typical adult is subject to "RT". A critical feature of "RT" is of
transdiagnostic significancefor example obsessive compulsive
disorder, insomnia and addictions involve unconstructive RT.
We argue that RTcannot be understood in isolation but must
rather be considered within models of whole minds. Researchers
must adopt the designer stance in the tradition of Artificial
Intelligence augmented by systematic conceptual analysis [2].
This means developing, exploring and implementing cognitive-
affective architectures. Empirical research on "RT" needs to be
driven by such theories, and theorizing about “RT” needs to
consider such data. We draw attention to H-CogAff theory of
mind (motive processing, emotion, etc.) and a class of emotions
it posits called perturbance (or tertiary emotions) [3,4], as a
foundation for the research programme we advocate. Briefly, a
perturbance is a mental state in which motivators tend to disrupt
executive processes. We argue that grief, limerence (the
attraction phase of romantic love) and a host of other
psychological phenomena involving "RT" should be
conceptualized in terms of perturbance and related design-based
constructs. We call for new taxonomies of "RT" in terms of
information processing architectures such as H-CogAff. We
claim general theories of emotion also need to recognize
perturbance and other architecture-based aspects of emotion.
Meanwhile “cognitive” architectures need to consider
requirements of autonomous agency, leading to cognitive-
affective architectures.
In the evenings she peeped out at him from the bookcase, from
the fireplace, from the corner he heard her breathing, the
caressing rustle of her dress. In the street he watched the
women, looking for someone like her. (“The Lady with the Dog”,
Anton Chekhov.)
Over 35 years ago, Aaron Sloman and Monica Croucher
launched a research programme based on an important , subtle
insight that had first been suggested by Herbert Simon [61]:
Humans have emotions for the same reason that future robots
will, as a consequence of interacting information processing
mechanisms that address the requirements of autonomous
agency [5,6]. This theory was grounded in Artificial Intelligence
1 Faculty of Education, Simon Fraser Univ. Email:
2 Dept. of Psychology, Univ. of Bath. Email:
3. Psychometrix Associates, Inc. Email:
(AI) and conceptual analysis. From 1990 to 2005 the ensuing
Cognition and Affect (CogAff) project (mainly at the University
of Birmingham, England) actively pursued this insight, by
exploring, implementing, assessing and refining requirements,
and software tools, and developing cognitive-affective
architectures capable of modelling the hypothesized affective
Sloman ultimately proposed three major types of emotion:
primary, secondary and tertiary [3]. Tertiary emotion, the focus
of [5,6] where it is simply called emotion, is also the focus of
this paper; for reasons explained below, we refer to it as
perturbance [4]. In a nutshell, in order for a motive to disturb
deliberative processes, such as problem solving, it must
implicitly or explicitly be assigned a sufficiently high insistence
level. A perturbance is a state in which an insistent motivator
tends to distract or alter deliberative processes in a manner that
is difficult for reflective processes to suppress or control. These
terms are briefly described below, and more extensively in
various CogAff project publications we cite. In short, the
concept of perturbance provides a parsimonious, design-based
way of understanding the obsessive aspects of emotion-like
states, wherein the agent experiences a certain loss of control of
attention and hence of management processes.
In this paper, we argue that perturbance is a major feature of
the human mind that has been ignored by psychologists but
deserves considerable attention. This concept has the potential to
unify several areas of study, including fundamental processes
such as attention, emotion and emotion regulation, cognitive
phenomena such as intrusive thought, and psychopathological
conditions such as rumination, obsessive worrying and
addictions. Like any theoretical concept, the concept of
perturbance does not stand alone. It is meaningful, promising
and useful because of the theoretical framework within which it
is embedded: a) the CogAff architecture schema, and b) H-
CogAff, a particular architecture based on CogAff which is
aimed specifically at understanding humans [3].
Whereas Sloman made significant attempts to disseminate the
design-based approach and H-CogAff to emotion and AI
researchers, the impact on the psychology literature so far has
been minimal, due to various factors some of which we will
allude to here. Meanwhile, affective computing (AC), a
discipline of computer science that focuses on emotion,
including emotion modelling, is gaining momentum. However,
AC currently tends to pursue narrow problems relevant to
practical applications focusing on primary emotions (e.g.,
machine perception of primary emotions). In AC, there is almost
no research on automatically detecting perturbance, let alone
attempts to produce systems that can experience and monitor
perturbance. This is the case despite the fact that Sloman’s work,
including the concept of perturbance, was described in Rosalind
Picard influential Affective Computing [65]. Sloman, who was
one of the first AI researchers to systematically emphasize
computational architectures, did foresee that AC would be a
long road [66]. Still, AI’s highly visible progress, and its work
on architectures, bode well for AC. We believe that history will
prove Sloman’s theory of perturbance is a sleeping beauty.
According to [8] these beautiestend to awakenwhen they
are discovered by a new community of researchers.
This paper is meant to promote consideration of H-CogAff by
indicating its relevance to many phenomena and research
communities, while focusing on one of its original concepts,
perturbance. However, we only have space for a cursory
overview of the theory itself. For more information about it, see
[2-7,9-10] and other papers cited below.
Sloman & Croucher [5,6] claimed emotions will emerge as side-
effects in minds designed to meet the requirements of
autonomous agency. These challenges include dealing with
multiple endogenous sources of motivation with limited physical
and processing resources in a rapidly changing, unpredictable,
and only partially controllable environment. Autonomous agents
require relatively simple mechanisms to generate and activate
goals. For various a priori reasons their deliberative mechanisms
have limited parallelism (see [4] ch. 4 and [61]) .
Not every activated goal can be considered simultaneously by
deliberative processes. There must be comparatively simple
mechanisms to decide whether the deliberative layer of the
architecture may be interrupted (or otherwise influenced) by a
given goal. These include insistence assignment, which
heuristically reflects the importance and urgency of a goal, and
interrupt filtering. For example, if a hungry autonomous agent
detects a rare opportunity to consume a source of energy, a new
goal to approach the source may be triggered. However, in order
for this goal to even be considered, it needs to be sufficiently
insistent to penetrate the attention filter and interrupt current
executive processing and behaviour. If the agent is under attack,
its executive processes might not even notice its goal to
approach the source because the filter threshold will have been
raised higher than the insistence level of the goal to approach. As
any good software designer knows, designing software involves
trade-offs. It’s impossible to design perfect insistence and
filtering rules. Sometimes, the robot will tend to be distracted by
its own insistent goals that it keeps rejecting (e.g., to approach an
appealing agent the pursuit of whom would violate its norms or
other goalsconflicted robot love.) Thus, not all emotion-like
states need be built into a robot; perturbant emotions will
Sloman, Beaudoin and their colleagues on H-CogAff project
continued to be challenged by psychologists who insisted that
the states they were describing were not really emotions.
Meanwhile, psychologists still do not agree on the meaning of
the word “emotion” [11-13], a highly polymorphous concept in
ordinary language. In order to avoid pointless turf wars over a
label and to stimulate progressive research, Beaudoin [4] coined
the term “perturbance” for Sloman’s original technical concept
of emotion [5]. Since then, one has been able to say there are
perturbant emotions (or perturbant states), while allowing
researchers to stipulate other types of emotion. We prefer the
term perturbance to ‘tertiary emotion’ (introduced later by
Sloman [3]) because the former denotes a more general
concepte.g., it can be interpreted in terms of architectures
with more than three layers.
Perturbance is of considerable adaptive significance because
it is an affection of the human brain’s executive processes, which
govern the agent.
Alas, internal attentional disturbance still does not figure
prominently in general theories of human emotion (e.g., [13,14]).
Ironically, it is in a biological theory of emotion that such
disturbance is highlighted, in what Panksepp & Biven [15] also
call tertiary emotions. Unfortunately, the architecture-based
concept of perturbance is still not used widely outside H-
CogAff. Yet the loss of control of attention of many emotional
episodes needs to be accounted for in such terms. We believe
that the concept of perturbance and its label still need to be
disseminated. It is our hope that this paper will help the idea gain
acceptance and treat this alexithymia in the literature on affect.
The concept of perturbance is part of a design-based research
programme that proposes a class of mental architectures
(CogAff) whose particular instance, H-CogAff, is the backdrop
of this paper [7]. H-CogAff is a response to human autonomous
agency requirements emanating from that programme. They
were alluded to above, and elaborated in [4]. A sketch of H-
CogAff is presented next to the more generic CogAff schema in
Figure 1.
Figure 1. CogAff schema and H-CogAff architecture diagrams
from [3]
This highly interconnected architecture assumes mechanisms
for perceiving and affecting the environment, generating alarms,
and creating and activating goals (and other types of motivators)
in real-time, synchronously and asynchronously from executive
processes. There are two broad types of executive processes:
deliberative processes (manifold base level reasoning,
evaluating, planning, scheduling, deciding, and control, also
known as management processes) and meta-management
processes (reflection and high-level control), localized in the
upper two levels of the H-CogAff architecture. The meta-
management layer could, for instance, postpone the
consideration of a newly activated goal till some juncture
deliberation scheduling. The “reactive” layer is more closely
coupled to the environment than the other two. H-CogAff
supposes variable-threshold qualitative and quantitative interrupt
filters, which protect limited-capacity executive processes. A
detailed specification of the structure of goals and the H-CogAff
processes operating on them is provided in [4].
On the basis of this architecture, Sloman was able to
distinguish three types of emotion [3,16]. Primary emotion
involves alarms triggered by perceptual information, such as an
angry glare or the unexpected appearance of the object of one’s
infatuation. Secondary emotion involves alarms triggered by
noticing an executive layer’s content (e.g., suddenly realizing a
plan of action will or would have a disastrous side-effect).
Alarms have global effects in the architecture, physiological or
exclusively mental. Tertiary emotionmore generally,
perturbanceinvolves an interaction, between motive
activators, filters, deliberative and meta-management processes,
etc. In perturbance, even if the deliberative layer were to
postpone consideration of an insistent motivator, the motivator
would still tend to penetrate the filter and divert deliberative
processes; or the motivator might maintain control of executive
processes. Perturbance is an emergent phenomenon; special
cases aside, it is not necessarily adaptive or maladaptive.
Adaptiveness and function are attributes of the architecture and
its constituent mechanisms.
The potential of this theory for psychology derives partly
from the research methods, the designer stance [67], that gave
rise to it. This stance can address a deep issue that surrounds but
has not previously been explicitly linked to psychology’s
“replication crisis” [17]. Psychology often lacks sufficient theory
for the phenomena it empirically investigates. We call for (1) a
better explicit characterization of human capabilities, an
exploration of mental architectures (designs), and
implementations [2]; and (2) empirical research driven by
unified theories of mind [18,19]. Cognitive architectures, still
largely ignored in psychology, are not enough; affective
processes deserve equal consideration. H-CogAff is still
incomplete; but it is a starting point worth considering.
Let us consider two emotions that can last for months, are
eminently perturbant, but are insufficiently explained by general
theories of emotion: grief and limerence. Perhaps this oversight
is because these emotions cannot ethically be manipulated in the
laboratory. Empirical psychology needs to be more concerned
with explaining observed individual possibilities in detail
(Newell, 1973), including detailed case studies, diary studies,
correspondence studies (e.g., [20]) and fiction ([21-23]).
Researchers could do worse than to try to design minds that
support the perturbant emotions depicted in humanity’s greatest
works of romantic poetry and fiction such as of the main
characters in Shakespeare’s Romeo & Juliet and Chekhov’s The
Lady with the Dog.
Grief. When grieving, one tends to be assailed by memories
and motives about the deceased. Wright, Sloman & Beaudoin
[24,25] offered a design-based explanation of emotion,
illustrated by a case study of grief, in which they claimed grief is
(often) an extended process of cognitive reorganization
characterized by the occurrence of negatively valenced
perturbant states caused by an attachment structure reacting to
news of the death.” That theory addresses important questions
such as: Why does grief consume the mourner? Because
executive processes have limited capacity and become swamped
by highly insistent motivators generated by a structure of
attachment to a highly valued individual; in addition, re-learning
and detachment require extensive rumination, which can
maintain perturbance.
Limerence. The prototypical perturbant state is limerence:
The nearly universal attraction phase of romantic love [26,27].
Notably, limerence researchers agree that limerence is
characterized by focused attention on and intrusive mentation
(IM) about the limerent object (LO) with manifold intense and
insistent motives for union with the LO ([26]). Limerence is of
great evolutionary and human significance, because it enhances
the likelihood of matingand, in most cultures, attaching to the
LO, which helps offspring survive [28]. Yet affective scientists
have hardly considered the phenomenon as a generally
representative and illuminating emotion, let alone from the
designer stance.
A defining feature of perturbance is diminishment of the
already limited human capacity to control one’s own attention.
Consider a limerent’s diary entry ““This obsession has infected
my brain. I cannot shake those constantly intruding thoughts of
you. Every thought winds back to you no matter how hard I try
to direct its course in other directions.”” [26]). Many, perhaps
most, limerent minds are aware of this intrusiveness. This is only
possible because (unlike most species) humans can, to a limited
extent, monitor and voluntarily control their attention (i.e.,
execute meta-management functions).
The H-CogAff framework seems to be at least as promising
for limerence as it is for grieftwo emotions that normally
operate in opposite ways on attachment structures. Limerence,
the attraction phase, involves establishing attachment structures:
motivators, motive generators, insistence assignment rules, other
reactive processes, filters, plans, etc. Grief is an extended
process of dismantling such attachment structures. Limerence
and grief overlap in heartbreak and lovelornness. Also like grief,
limerence can loosen prior attachment (facilitating the
abandonment of one’s current partner for a new one, or
forgetting a prior love). Accounting for attachment processes is
important given that emotions seem to have evolved in large part
to enable individuals to indirectly manage each other via
commitments and attachments [29]. Several H-CogAff projects
have already examined perturbance in relation to attachment
(e.g., [30]).
While it may be tempting to cast limerence as a pathological
form of romantic love [31,32], this would distort the original and
common academic conception of limerence [33]. This would
also overlook the near universality and evolutionary significance
of limerence. Like other obsessions and other emotional states,
limerence lies on continua [34] and may or may not be
pathological. We believe the distorted casting should be resisted
by scholars; instead other terms should be used to describe
pathological limerence. We also recommend that scientific
literature on this phase converge on the term limerenceand
help shape folk psychology.
There is more to limerence than perturbance, just as there is
more to motive processing and emotions than perturbance.
Perturbance is a particularly promising concept partly because
it encourages questions to be raised progressively about mental
states in terms of whole-mind design (motive generators,
attachment structures, etc.), leading to further requirement and
design specification. Perturbance cannot be understood in
isolation. It transcends folk psychology and the intentional
Watkins [1] suggested that an important attentional
phenomenon should be conceptualized as “repetitive thought”
(RT). He echoed a definition of RT as a “process of thinking
attentively, repetitively or frequently about one’s self and one’s
world [forming] the core of a number of different models of
adjustment and maladjustment.” (p. 163) Under the banner of
RT, Watkins included such varied phenomena as cognitive and
emotional processing of persistent intrusions, depressive
rumination, perseverative cognition, rumination, worry,
planning, problem solving, and mental simulation, mind
wandering, counterfactual thinking, post-event rumination,
defensive pessimism, positive rumination, reflection, habitual
negative self-thinking. To this list we would add obsessive and
compulsive mentation and cravings. Watkins notes that worry,
for instance, was defined in [35] as “a chain of thoughts and
images, negatively affect-laden and relatively uncontrollable”
and as “an attempt to engage in mental problem-solving on an
issue whose outcome is uncertain but contains the possibility of
one or more negative outcomes” (p. 9).
Watkins’s reasons for favouring RT as the overarching
concept were that it is more inclusive than the alternatives,
atheoretical, clearer, highly correlated with measures of worry
and rumination, and non-evaluative (constructive or
We agree that RT phenomena are scientifically significant.
For many of them are typical of normal self-regulation
everyone experiences IM, for instance. Furthermore, extreme
forms of RT are transdiagnostic [36]. RT plays a critical role in
insomnia and depression, for instance [37]. Insomnia also is of
transdiagnostic significance [37]a cause and consequence of
However, Watkins’ RT conceptualization is limited. Firstly,
the expression “RT” misleadingly suggests that the repetitive
content is cognitive in the traditional sense (“thought”), whereas
it is often affectively-laden. Moreover, the processes that
manage the ‘repetitive’ mental content serve these motivators,
such as assessing and deciding. Repetitive mentation (RM) is
more inclusive and germane. Further, the atheoretical criterion is
unrealistic and counterproductive; it also runs against Watkins’s
other criterion of being conceptually clear. One needs a general
theory, beyond folk psychology, in relation to which intrusions
and the executive processes that respond to them are specified.
(Compare progress in evolutionary classification based on
molecular genetics rather than phenotypic features.) Whether
authors are explicit and clear or not about their theory, the
concepts at play in RT involve, or at least require, a functional
architecture. For something must be generating motivators;
something must be interrupting in intrusions; something must be
considering goals; something must be prioritizing them; etc.
These mechanisms need to be named and specified in relation to
an architecture. The theory ought to “cut nature at its joints” and
be amenable to a progressive research programme of simulation,
further theoretical development and cumulative empirical
research [38]. Furthermore, the all inclusive RT
conceptualization comes at the cost of papering over significant
differences, for instance between reflection and rumination. The
farrago of “RT” concepts requires conceptual analysis and
functional specification, which will lead to much pruning and
The phenomena of RM are too global, involving too many
diverse wide-ranging mechanisms of mind, to be understood
without reference to a broad and explicit theory of mind.
Moreover, one must understand the how of normal information
processing (IP) to assess mentation as constructive or
unconstructive. Alas, the RT literature has failed to adopt or
develop architectural models of mind. For instance, in describing
a highly studied phenomenon of RM, affective biases, Mathews,
Mackintosh & Fulcher [39] invoke interrupt signals, attentional
vigilance, effortful suppression and intrusions. The concepts of
cognitive and attentional ‘biases’ [68], are currently cast mainly
in terms of ‘external and internal stimuli’ rather than in terms of
goal or motive processing (contrast [4-5,61]), i.e., the
mechanisms that are being affected. The attentional bias and RM
literatures fail to invoke an overall model of mind which, for
instance generates motives, prioritizes, them and acts upon them,
i.e., that addresses the types of capabilities with which H-CogAff
is concerned.
Wells & Mathews published a book length theory, the Self-
Regulatory Executive Function (S-REF) model [40], that
valiantly attempts to address many phenomena at the intersection
of cognition and affect, including RT. The model is explicitly
inspired by architecture-based AI. However, the empirical RT
literature seems at most to pay lip service to it. For instance, in
their extensive book on transdiagnostic processes, Harvey et al.
[36] summarily reject S-REF. It is not noted that and how S-REF
would need to be improved to address more of the requirements
of autonomous agency normal multi-purpose (multi-motive)
competence. The main issue, that this promising underdeveloped
theory needed AI attention was not mentioned.
Watkins (2008) and others point to control theory as an
explanatory framework for “RT” and self-regulation. While
some of these models are promising (e.g., [41]), they too need to
be integrated with a broader architecture. They need to deal with
rich qualitative control states and mechanisms that follow from
the requirements of autonomous agency (see [42]).
H-CogAff provides a theoretical framework in relation to
which classification and modelling may proceed. This
framework has the advantage of being constructed to explore
how human minds might solve real world problems of
autonomous agency. It is by no means a complete or detailed
specification; but it has proven to be useful for generating and
exploring models, many of which have already been
implemented [7,9].
H-CogAff offers a path towards a deeper conceptualization of
“RT”. In [1], intrusive thought (IT) is not a category of RT,
likely because it is an essential aspect of RT. IT is better, and
more generally, conceived as intrusive mentation (IM), and more
deeply as perturbance. The concept of perturbance is based on
the dispositional concept of insistence of mental content: a
motivator may be insistent and yet not disrupt processing. To
understand IM as perturbance we must specify in terms of an
architecture (like H-CogAff) the ways in which insistence
assignment, interrupt filtering and attention switching are
This may also help address the need in the RT literature for a
design-based taxonomy of patterns of executive processes. [4]
and [25] put forth several categories, such as oscillation between
decisions, manifest perturbance, digressions and maundering.
Several other patterns have been identified in the CogAffect
project (e.g., [25,43]). These, and several types of phenomena
labelled by Watkins as RT (such as worry and rumination) need
to be systematically characterized in terms of patterns of
interaction between management, reflective and reactive
processes in H-CogAff.
Several other research problems need to be reinterpreted
specifically in terms of perturbance and, more generally, from
the designer stance. Motivation in psychology tends to be
conceived as the directing and energizing of behaviour [44]
(what goals do people choose; when, why, and how intensely do
they pursue them), rather than in terms of motive processing
(how can motives be processed to evince autonomous agency).
For instance, none of the Behavior & Brain Sciences peer
responses to the Selfish Goal theory [45] noted its lack of
explicit architecture nor that its goal specification and processes
are bare (e.g., where is insistence? Contrast [4]). Pleasure and
avoidance of pain are still normally assumed to be the final ends,
while the deeper, more subtle and generative possibility of
architecture-based motivation [46] is ignored even in rare
discussions of effectance ([41]; contrast [47]). Stanovich
developed a promising theory [48] to explain and improve
rationality with a three-level architecture which, although
referring to H-CogAff, fails to use motive processing constructs.
Yet the perturbance theory was meant to account for breakdowns
in rationality [5]. Meanwhile, the recent theory of cognitive
energetics [49], which is meant to explain all instances of goal-
directed thinking, also lacks an architecture (contrast the related
concept of economy of mind in Wright [25]).
Given that perturbance is an underlying construct to explain
RT, and RT is transdiagnostic, it stands to reason that the
concept of perturbance is relevant to transdiagnostic approaches.
For instance, addictions involve motivators that are both insistent
(attention grabbing) and intense (control behaviour). Obsessions
and compulsions also involve perturbance. More generally, a
design-based approach is required for transdiagnostic
understanding [50]. Even more generally, to understand
abnormal psychology we must understand normal psychology in
design-based terms.
Pain in its various forms involves aversive perturbance and
should be modelled with H-CogAff or related designs.
Beaudoin [47] argued that mindfulness-based therapies,
which are either explicitly behaviourist [51] or use architectures
detached from AI, could benefit from H-CogAff. Mindfulness
therapies assume direct experience [51]. But no one has ever
built a machine that can directly perceive anything, nor
demonstrated the possibility of such a machineperception is in
fact always highly indirect. Mindfulness therapies prescribe
awareness of emotion, but by this term their authors mainly refer
to affective feelings. Shouldnt therapists and clients be trained
with a rich design-based theory of mind to improve clients’
awareness, i.e., models of themselves? Similarly, the acceptance
and commitment therapy (ACT) technique of “cognitive
defusion” [51] requires an IP ontology of mental states that ACT
fails to invoke.
Perturbance is also quite relevant to human memory.
Following Anderson’s adaptive explanation of memory [62],
Beaudoin [47] proposed the heuristic relevance-signaling
hypothesis (“HRS”) from the designer stance. On a daily basis,
humans process enormous amounts of information. The brain
cannot deeply interpret it all, nor store all of its interpretations.
Nor can the cortex directly signal relevance top down (The direct
command “I shall remember this phone number” does not work.)
What information should be given precedence? Testing effects
are amongst the most well documented findings in empirical
psychology: repeatedly recalling information potentiates it. The
HRS hypothesis states that deliberative layer recall attempts are
implicit cues to the brain’s heuristic memory indexing
mechanisms to prioritize access to information (‘memories’)
related to the perturbanceinformation (interpretations,
narratives, etc.) that the deliberative layer has at least attempted
to recall (reconstruct). Perturbances are hijackings of these
mechanisms by insistent motivators, potentiating memories
related to the perturbant objects (e.g., the limerent object).
Psychology has struggled with the question: in what respect
can the experience of music in particular and art more generally
be emotional? From the designer stance we might similarly ask
how can great art rivet us and reverberate within us, from catchy
ear worms to more? We suggest a new answer based on H-
CogAff theory, namely that music and fiction may trigger an
illusion of perturbance: the reflective-layer impression that the
agent is experiencing a genuine perturbance (as if self-generated
motives were insistently being activated, captivating
management processes). More obviously, art likely often
operates by increasing the insistence of one’s own latent
motivators (triggering limerence and grief, for instance). To
explore and specify these vague hypotheses, we suggest
modeling responses to high-calibre, multi-modal art depicting
limerence and grief that uses repetition in provocative ways,
such as Veda Hill & Amiel Gladstone’s musical theatre
adaptation of Tchaikovsky opera, Onegin [64], itself based on
Pushkin’s poem, Eugene Onegin.
It should be noted that perturbance is not the only type of loss
of control in minds. Dean Petters described several other types in
relation to H-CogAff [43].
We also believe a theory of perturbance can be used for
positive psychology and self-help. For example, Beaudoin
(2013) developed the cognitive shuffle a technique to combat
insomnia which is meant to work partly by interfering with
bedtime perturbance [52]. Focusing and flow are essential to
cognitive productivity and hence to knowledge economies.
Distraction is largely affective yet theories of attention and
knowledge translation on the subject e.g. [53-54] Levitin (2014),
Gallagher (2006) do not deal with motive processing and fail
to invoke perturbance. Theories of learning, expertise and
productive practice need to explain how humans can deliberately
develop their mental architectures, e.g., creating new goal
generators [47,55-56].
In short, previous research phenomena and problems can
systematically be revisited from the designer stance as involving
We have called attention to perturbance as a way to understand a
broad variety of normal and pathological mental phenomena in
IP terms. This concept has the advantage of being firmly rooted
in AI and of involving a flexible, extensible architectural
framework. This enables research problems to be considered in
terms of models of entire minds.
Perturbance and other aspects of H-CogAff are not final
explanations. They are part of the beginning of what we believe
can be a progressive research programme.
The designer stance also is directly relevant to education and
training. Psychology students need to be able to think about
themselves, other humans and possible minds in terms of
multiple cognitive-affective IP architectures. Psychology and AI
students should also graduate well-trained in conceptual
analysis [57,58] as they are in empirical research methods.
(These would be fitting topics in [59], for example.)
We are not suggesting a one-way flow of influence. Instead,
we advocate a progressive theory-driven research programme to
improve H-CogAff and related proposals. There is a need for
more AI researchers to consider broad, integrative, multi-
layered, affective autonomous agency. We believe psychology
and AI researchers need to work more closely together, not only
on purely cognitive problems but affective ones as well. AI and
psychology must blend more. For the opening quotation of
Beaudoin’s (1994) Ph.D. thesis [4] is still true: “The problem is
not that we do not know which theory is correct, but rather that
we cannot construct any theory at all which explains the basic
facts” [60] (p. 109.)
We would like to thank Dr. Al Sather, Carol Woodworth and
two referees for their comments which helped improve this
[1] E. R. Watkins, “Constructive and unconstructive repetitive
thought,” Psychological Bulletin 134, 163206 (2008).
[2] A. Sloman, “Prospects for AI as the general science of
intelligence,” 1993, Amsterdam, 110, IOS Press.
[3] A. Sloman, “How many separately evolved emotional beasties
live within us?,” in Emotions in humans and artifacts, R. Trappl,
P. Petta, and S. Payr, Eds. (MIT Press, 2003).
[4] L. P. Beaudoin, “Goal processing in autonomous agents”
(Birmingham, England, 1994).
[5] A. Sloman and M. Croucher, “You don't need a soft skin to have a
warm heart: Towards a computational analysis of motives and
emotions,” 004, 1981.
[6] A. Sloman and M. Croucher, “Why robots will have emotions,”
[7] A. Sloman, “The Cognition and Affect project: Architectures,
architecture-schemas, and the new science of mind,” 2008.
[8] Q. Ke, E. Ferrara, F. Radicchi, and A. Flammini, “Defining and
identifying Sleeping Beauties in science,” Proceedings of the
National Academy of Sciences 112, 74267431 (2015).
[9] N. Hawes, “A survey of motivation frameworks for intelligent
systems,” Artificial Intelligence 175, 10201036 (2011).
[10] E. Hudlicka, “Affective BICA: Challenges and open questions,”
Biologically Inspired Cognitive Architectures 7, 98125 (2014).
[11] T. Read and A. Sloman, “The terminological pitfalls of studying
emotion,” 18 (1993).
[12] C. E. Izard, “The many meanings/aspects of emotion: Definitions,
functions, activation, and regulation,” Emotion Review 2, 363370
[13] J. A. Russell, “Emotion, core affect, and psychological
construction,” Cognition & Emotion 23, 12591283 (2009).
[14] K. R. Scherer, “What are emotions? And how can they be
measured?,” Social Science Information 44, 695729 (2005).
[15] J. Panksepp and L. Biven, “The Archaeology of Mind:
Neuroevolutionary Origins of Human Emotions” (2012).
[16] A. Sloman, R. Chrisley, and M. Scheutz, “The architectural basis
of affective states and processes,” in Who needs emotions? The
brain meets the robot, J. M. Fellous and M. A. Arbib, Eds. (New
York: Oxford University Press, 2005).
[17] S. E. Maxwell, M. Y. Lau, and G. S. Howard, “Is psychology
suffering from a replication crisis? What does ‘failure to replicate’
really mean?,” The American psychologist 70, 487498 (2015).
[18] A. Newell, Unified theories of cognition (Harvard University
Press, Cambridge, MA, 1990).
[19] A. Wells and G. Mathews, Attention and Emotion: A Clinical
Perspective (Lawrence Erlbaum Associates Publishers, Hillsdale,
NJ:, 1994).
[20] L. Nys, “Emotional ‘counter-practices’ in the discipline section of
the state re-education institution for female juvenile delinquents
(1927-1939),” 30 July 2015, 116.
[21] K. Oatley, Such stuff as dreams: The psychology of fiction (2011).
[22] K. Oatley, Best Laid Schemes (Cambridge Univ Press, Cambridge,
[23] P. C. Hogan, What Literature Teaches Us about Emotion
(Cambridge University Press, 2011).
[24] I. Wright, A. Sloman, and L. P. Beaudoin, “Towards a design-
based analysis of emotional episodes,” Philosophy, Psychiatry &
Psychology 3, 101126 (1996).
[25] I. P. Wright, “Emotional Agents” (1997).
[26] D. Tennov, Love and Limerence (Scarborough House, 1979).
[27] S. E. Reynolds, “‘Limerence’: A new word and concept,”
Psychotherapy 20, 107111 (1983).
[28] H. E. Fisher, “Lust, attraction, and attachment in mammalian
reproduction,” Human Nature 9, 2352 (1998).
[29] M. Aubé, “Unfolding commitments management: A systemic
view of emotions,” in Handbook of research on synthetic
emotions and sociable robotics New applications in affective
computing and artificial intelligence, J. Vallverdú and D.
Casacuberta, Eds. (New York, NY, 2009).
[30] D. Petters and L. P. Beaudoin, “Attachment modelling: From
observations to scenarios to designs,” in Computational
Neurology and Psychiatry, P. Erdi, B. S. Bhattacharya, and A.
Cochran, Eds. (2017).
[31] A. Wakin and D. B. Vo, “Love-variant: The Wakin-Vo IDR
model of limerence,” 2008.
[32] M. Reynaud, L. Karila, L. Blecha, and A. Benyamina, “Is Love
Passion an Addictive Disorder?,” The American Journal of Drug
and Alcohol Abuse 36, 261267 (2010).
[33] H. van Steenbergen, S. J. E. Langeslag, G. P. H. Band, and B.
Hommel, “Reduced cognitive control in passionate lovers,”
Motivation and Emotion, 444450 (2013).
[34] E. Hatfield and S. Sprecher, “Measuring passionate love in
intimate relationships,” Journal of Adolescence 9, 383419
[35] T. D. Borkovec, E. Robinson, and T. Pruzinsky, “Preliminary
exploration of worry: Some characteristics and processes,”
Behaviour Research and Therapy, 916 (1983).
[36] A. G. Harvey, Cognitive Behavioural Processes Across
Psychological Disorders (Oxford University Press, USA, 2004).
[37] M. R. Dolsen, L. D. Asarnow, and A. G. Harvey, “Insomnia as a
transdiagnostic process in psychiatric disorders,” Curr Psychiatry
Rep 16, 471 (2014).
[38] R. P. Cooper, “The role of falsification in the development of
cognitive architectures: Insights from a Lakatosian analysis,”
Cognitive Science 31, 509533 (2007).
[39] A. Mathews, B. Mackintosh, and E. P. Fulcher, “Cognitive biases
in anxiety and attention to threat,” Trends in cognitive sciences 1,
340345 (1997).
[40] A. Wells and G. Matthews, Attention and Emotion (Psychology
Press, 1995).
[41] O. Nafcha, E. T. Higgins, and B. Eitam, “Control feedback as the
motivational force behind habitual behavior,” in Motivation -
Theory, Neurobiology and Applications 229 (Elsevier, 2016).
[42] A. Sloman, Beyond turing equivalence. Red. P. Millican, A. Clark.
Machines And Thought: The Legacy Of Alan Turing, vol I: 179-
219 (1990).
[43] D. Petters, “Losing control within the H-Cogaff architecture,” in
From animals to robots and back: Reflections on hard problems
in the study of cognition 22 (Springer International Publishing,
Cham, 2014).
[44] K. Danziger, Naming the Mind (SAGE, 1997).
[45] J. Y. Huang and J. A. Bargh, “The Selfish Goal: Autonomously
operating motivational structures as the proximate cause of human
judgment and behavior,” The Behavioral and Brain Sciences 38,
121135 (2015).
[46] A. Sloman, “Architecture-based motivation vs. reward-based
motivation,” 2015.
[47] L. P. Beaudoin, Cognitive productivity: Using knowledge to
become profoundly effective (CogZest, Pitt Meadows, BC, 2014).
[48] K. E. Stanovich, Rationality and the reflective mind (Oxford
University Press, USA, 2011).
[49] A. W. Kruglanski, J. J. Bélanger, X. Chen, C. Köpetz, A. Pierro,
and L. Mannetti, “The energetics of motivated cognition: A force-
field analysis.,” Psychological Review 119, 120 (2012).
[50] E. Hudlicka, “Computational modeling of cognitionemotion
interactions: Theoretical and practical relevance for behavioral
healthcare,” in Handbook of Affective Sciences in Human Factors
and HCI, M. P. Jeon, Ed. (Elsevier, Waltham, MA, 2017).
[51] S. C. Hayes, K. D. Strosahl & K. G. Strosahl. Acceptance and
commitment therapy: The process and practice of mindful change,
Guilford Press, New York, 2011.
[52] L. P. Beaudoin, N. Digdon, and K. O'Neill, “Serial diverse
imagining task: A new remedy for bedtime complaints of
worrying and other sleep-disruptive mental activity,” 2016, A209.
[53] D. J. Levitin, The organized mind: Thinking straight in the age of
information overload (2014).
[54] W. Gallagher, Rapt (Penguin, 2009).
[55] P. H. Winne, “Self-regulated learning viewed from models of
information processing,” in Self-regulated learning and academic
achievement: Theoretical perspectives, 2nd ed., B. J. Zimmerman
and D. H. Schunk, Eds. (Lawrence Erlbaum, Mahwah, NJ, 2001).
[56] L. P. Beaudoin, “Developing expertise with objective knowledge:
Motive generators and productive practice,” in From Robots to
Animals and Back, J. Wyatt and D. Petters, Eds. (Springer, 2014).
[57] A. Sloman, The computer revolution in philosophy: Philosophy,
science and models of mind (Harvester Press, 1978).
[58] A. Ortony, G. L. Clore, and M. A. Foss, “The referential structure
of the affective lexicon,” Cognitive Science: A Multidisciplinary
Journal 11, 341364 (1987).
[59] K. E. Stanovich, How to think straight about psychology, 9 ed.
(Allyn & Bacon, 2004).
[60] R. Power, “The organisation of purposeful dialogues,” Linguistics
17, 107152 (1979).
[61] H. A. Simon. "Motivational and emotional controls of cognition"
Psychological Review, 74, 2939 (1967).
[62] Anderson, John R. "Is human cognition adaptive?" Behavioral and
Brain Sciences 14, 471-485 (1991).
[63] P. N. Johnson-Laird & K. Oatley "Emotions, Music, and
Literature" in Handbook of Emotions L. F. Barrett, M. Lewis & J.
M. Haviland-Jones (Eds). (New York, 2008).
[64] V. Hill & E. Gladstone. Onegin. (2016)
[65] R. W. Picard. Affective Computing. (MIT Press, 2000).
[66] A. Sloman. "Review of: Rosalind Picard’s affective computing."
AI Magazine, 20, 127-137 (1997).
[67] J. McCarthy. "The well-designed child." Artificial Intelligence,
172, 2003-2014 (2008).
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Recent years have witnessed an increasing interest in developing computational models of emotion and emotion–cognition interaction, within the emerging area of computational affective science. At the same time, emotion theorists and clinical psychologists have begun to recognize the importance of moving beyond descriptive characterizations of psychopathology, and identifying the underlying mechanisms that mediate both the etiology of affective disorders, and their treatment: the transdiagnostic approach to psychopathology. Computational models of cognition–emotion interactions have the potential to facilitate more accurate assessment and diagnosis of affective disorders, and to provide a basis for more efficient and targeted approaches to their treatment, through an improved understanding of the underlying mechanisms. This chapter discusses the state-of-the-art in modeling emotion–cognition interaction and the relevance of these models for understanding the mechanisms mediating psychopathology and therapeutic action. The discussion is limited to symbolic models and theories defined at the psychological, versus neural, level. The chapter also outlines how these models can support the development of serious therapeutic games, to enhance assessment and treatment methods in behavioral healthcare.
Full-text available
Motivated behavior is considered to be a product of integration of a behavior's subjective benefits and costs. As such, it is unclear what motivates 'habitual behavior' which occurs, by definition, after the outcome's value has diminished. One possible answer is that habitual behavior continues to be selected due to its 'intrinsic' worth. Such an explanation, however, highlights the need to specify the motivational system for which the behavior has intrinsic worth. Another key question is how does an activity attain such intrinsically rewarding properties. In an attempt to answer both questions, we suggest that habitual behavior is motivated by the influence it brings over the environment — by the control motivation system, including 'control feedback'. Thus, when referring to intrinsic worth we
Conference Paper
Full-text available
Introduction: A racing mind, worries, and uncontrollable thoughts are common bedtime complaints among poor sleepers. Beaudoin created a Serial Diverse Imagining task (SDIT) that can be used at bedtime to divert attention away from sleep interfering thoughts, An app randomly presents recordings of relatively concrete words one at a time with an 8-second interval between recordings during which the person creates and maintains a mental image of the word until the next recording prompts the next image and so on. Our study is an experimental test of SDIT compared to the standard treatment of Structured Problem-solving (SP) and to the combination of both treatments. A key feature of SP is that it must be done earlier than bedtime and requires about 15 minutes to do it. SDIT, which is done at bedtime, does not have those constraints. Method: 154 university students (137 female) who complained of excessive cognitive pre-sleep arousal were randomly assigned to receive SDIT, SP, or both. At baseline, they completed Pre-Sleep Arousal Scale (Somatic and Cognitive), Sleep Quality Scale, Glasgow Sleep Effort Scale and Sleep Hygiene Index. Depending on the measure, participants redid it one week and/or one month after starting the intervention. (They also completed sleep diaries and appraisals of the interventions, which are omitted due to space). Results: Repeated measures ANOVAs indicated that cognitive and somatic pre-sleep arousal , sleep effort, and sleep quality improved significantly relative to baseline (p < .001; Partial η2 = .43 to .71) even though sleep hygiene worsened ( p < .001; Partial η2 = .23). The latter finding is not unexpected because the baseline was done at the start of the academic term before the onset of academic pressures. The fact that we found sleep and arousal improvements in this context are notable. Conclusion: Beaudoin’s Serial Diverse Imagining Task (SDIT) was as effective as Structured Problem-Solving (SP) in reducing pre-sleep arousal, sleep effort, and poor sleep quality. One advantage of SDIT is that it can be done at bedtime, unlike SP.
Full-text available
Psychology has recently been viewed as facing a replication crisis because efforts to replicate past study findings frequently do not show the same result. Often, the first study showed a statistically significant result but the replication does not. Questions then arise about whether the first study results were false positives, and whether the replication study correctly indicates that there is truly no effect after all. This article suggests these so-called failures to replicate may not be failures at all, but rather are the result of low statistical power in single replication studies, and the result of failure to appreciate the need for multiple replications in order to have enough power to identify true effects. We provide examples of these power problems and suggest some solutions using Bayesian statistics and meta-analysis. Although the need for multiple replication studies may frustrate those who would prefer quick answers to psychology's alleged crisis, the large sample sizes typically needed to provide firm evidence will almost always require concerted efforts from multiple investigators. As a result, it remains to be seen how many of the recently claimed failures to replicate will be supported or instead may turn out to be artifacts of inadequate sample sizes and single study replications. (PsycINFO Database Record
The purpose of the research programme detailed in this paper is to update the attachment control system framework that John Bowlby set out in his formulation of Attachment Theory. It does this by reconceptualising it as a cognitive architecture that can operate within multi-agent simulations. This is relevant to computational psychiatry because attachment phenomena are broad in scope and range from healthy everyday interactions to psychopathology. The process of attachment modelling involves three stages and this paper makes contributions in each of these stages. Firstly, a survey of attachment research is presented which focuses on two important attachment behavioural measures: the Strange Situation Procedure and the Adult Attachment Interview (AAI). These studies are reviewed to draw out key behavioural patterns and dependencies. Secondly, the empirical observations that are to be explained in this research programme are abstracted into scenarios which capture key behavioural elements. The value of behavioural scenarios is that they can guide the simulation design process and help evaluate simulations which are produced. Thirdly, whilst the implementation of these scenarios is still a work in progress, several designs are described that have been created and implemented as simulations. These include normative and non-pathological infant behaviour patterns observed across the first year of life in naturalistic observations and ‘Strange Situation’ studies. Future work is described which includes simulating dysfunctional infant behaviour patterns and a range of adult attachment behaviour patterns observed in the Adult Attachment Interview. In conclusion, this modelling approach is distinguished from other approaches in computational psychiatry because of the psychologically high level at which it models phenomena of interest.
“Literature offers a veritable treasure trove of wisdom and insights about the nature and manifestations of human emotions, yet emotion researchers have been slow to explore this exciting domain. This book represents a groundbreaking attempt to bridge the gap between scientific research and complementary literary insights on emotions. The chapters explore in considerable detail such core emotions as love, guilt, mirth, shame, and compassion, drawing on the work of such literary giants as Shakespeare. The author takes us on an exhilarating journey of discovery of the subtleties, structure, and functions of human emotions using an ingenious approach fusing art and science. This book will be warmly welcomed by all researchers, teachers, students, and professionals interested in understanding emotions, and will be enjoyed by everyone who is fascinated by the intricacies of human emotionality.” Joseph P. Forgas, University of New South Wales “In What Literature Teaches Us about Emotion, leading literary cognitivist Patrick Colm Hogan stages readings of well-selected literary texts illustrating love, grief, mirth, guilt, shame, jealousy, disgust, compassion, and pity. Beyond their thematic resonances, Hogan’s chosen texts serve as a source of knowledge about how human emotions work. Illuminating and suggestive for conversations in affective literary studies, this book lends itself to discussion in the classroom, where the dialogs about texts by means of which we come to understand our responses to the world and to one another take place.” Suzanne Keen, Washington and Lee University “What Literature Teaches Us about Emotion provides an extraordinarily lucid and insightful account of the relevance of the cognitive sciences to literary study, as well as the potential for literary studies to contribute to a genuinely interdisciplinary history of emotion.” Evelyn Tribble, University of Otago, New Zealand
The paper describes progress towards producing deep models of emotion. It does this by working towards the development of a single explanatory framework to account for diverse psychological phenomena where control is lost. Emotion theories are reviewed to show how emotions can be described according to various emotional components and emotional phases. When applied in isolation, these components or phases can result in shallow models of emotion. The CogAff schema and HCogAff architecture are presented as frameworks, which can be used to organise and integrate these various separate ways in which emotion can be described. The examples of losing control discussed in this paper include: short-term emotional interrupts to ongoing processing; experiencing grief after the loss of an attachment figure; longer term emotional strengthening of motivation; using strong emotions to guarantee keeping one's own commitments; Freudian Repression as a defensive loss of access to painful information; and self-deception as a general strategy in deceiving others.
This book attempts to resolve the Great Rationality Debate in cognitive science-the debate about how much irrationality to ascribe to human cognition. It shows how the insights of dual-process theory and evolutionary psychology can be combined to explain why humans are sometimes irrational even though they possess remarkably adaptive cognitive machinery. The book argues that to characterize fully differences in rational thinking, we need to replace dual-process theories with tripartite models of cognition. Using a unique individual differences approach, it shows that the traditional second system (System 2) of dual-process theory must be further divided into the reflective mind and the algorithmic mind. Distinguishing them gives a better appreciation of the significant differences in their key functions: the key function of the reflective mind is to detect the need to interrupt autonomous processing and to begin simulation activities, whereas that of the algorithmic mind is to sustain the processing of decoupled secondary representations in cognitive simulation. The book then uses this algorithmic/reflective distinction to develop a taxonomy of cognitive errors made on tasks in the heuristics and biases literature. It presents the empirical data to show that the tendency to make these thinking errors is not highly related to intelligence. Using a tripartite model of cognition, the book shows how, when both are properly defined, rationality is a more encompassing construct than intelligence, and that IQ tests fail to assess individual differences in rational thought. It then goes on to discuss the types of thinking processes that would be measured if rational thinking were to be assessed as IQ has been.