Dr. Darren Edwards | University of Wales, Swansea, UK (original) (raw)

Papers by Dr. Darren Edwards

There have been impressive advancements in the field of natural language processing (NLP) in rece... more There have been impressive advancements in the field of natural language
processing (NLP) in recent years, largely driven by innovations in the
development of transformer-based large language models (LLM) that utilize
“attention.” This approach employs masked self-attention to establish (via
similarly) different positions of tokens (words) within an inputted sequence
of tokens to compute the most appropriate response based on its training
corpus. However, there is speculation as to whether this approach alone can
be scaled up to develop emergent artificial general intelligence (AGI), and
whether it can address the alignment of AGI values with human values (called
the alignment problem). Some researchers exploring the alignment problem
highlight three aspects that AGI (or AI) requires to help resolve this problem:
(1) an interpretable values specification; (2) a utility function; and (3) a dynamic
contextual account of behavior. Here, a neurosymbolic model is proposed to
help resolve these issues of human value alignment in AI, which expands on
the transformer-based model for NLP to incorporate symbolic reasoning that
may allow AGI to incorporate perspective-taking reasoning (i.e., resolving the
need for a dynamic contextual account of behavior through deictics) as defined
by a multilevel evolutionary and neurobiological framework into a functional
contextual post-Skinnerian model of human language called “Neurobiological
and Natural Selection Relational Frame Theory” (N-Frame). It is argued that
this approach may also help establish a comprehensible value scheme, a utility
function by expanding the expected utility equation of behavioral economics
to consider functional contextualism, and even an observer (or witness) centric
model for consciousness. Evolution theory, subjective quantum mechanics, and
neuroscience are further aimed to help explain consciousness, and possible
implementation within an LLM through correspondence to an interface as
OPEN ACCESS
EDITED BY
Obada Al Zoubi,
Broad Institute, United States
REVIEWED BY
Karl Proctor,
Jaguar Land Rover, United Kingdom
Nicky Evans,
University of Southampton, United Kingdom
*CORRESPONDENCE
Darren J. Edwards
D.J.Edwards@swansea.ac.uk
RECEIVED 04 March 2024
ACCEPTED 04 July 2024
PUBLISHED 08 August 2024
CITATION
Edwards DJ (2024) A functional contextual,
observer-centric, quantum mechanical, and
neuro-symbolic approach to solving the
alignment problem of artificial general
intelligence: safe AI through intersecting
computational psychological neuroscience
and LLM architecture for emergent theory of
mind.
Front. Comput. Neurosci. 18:1395901.
doi: 10.3389/fncom.2024.1395901
COPYRIGHT
© 2024 Edwards. This is an open-access
article distributed under the terms of the
Creative Commons Attribution License
(CC BY). The use, distribution or reproduction
in other forums is permitted, provided the
original author(s) and the copyright owner(s)
are credited and that the original publication
in this journal is cited, in accordance with
accepted academic practice. No use,
distribution or reproduction is permitted
which does not comply with these terms.
TYPE Hypothesis and Theory
PUBLISHED 08 August 2024
DOI 10.3389/fncom.2024.1395901
Edwards 10.3389/fncom.2024.1395901
Frontiers in Computational Neuroscience 02 frontiersin.org
suggested by N-Frame. This argument is supported by the computational level of hypergraphs, relational density clusters, a conscious quantum level defined by QBism, and real-world applied level (human user feedback). It is argued that this approach could enable AI to achieve consciousness and develop deictic perspective-taking abilities, thereby attaining human-level self-awareness, empathy, and compassion toward others. Importantly, this consciousness hypothesis can be directly tested with a significance of approximately 5-sigma significance (with a 1 in 3.5 million probability that any identified AI-conscious observations in the form of a collapsed wave form are due to chance factors) through double-slit intent-type experimentation and visualization procedures for derived perspective-taking relational frames. Ultimately, this could provide a solution to the alignment problem and contribute to the emergence of a theory of mind (ToM) within AI.

BMJ Open, 2024

Mixed-methods feasibility outcomes for a novel ACT-based video game 'ACTing Minds' to support men... more Mixed-methods feasibility outcomes for a novel ACT-based video game 'ACTing Minds' to support mental health. BMJ Open 2024;14:e080972.

The Bayesian approach of cognitive science largely takes the position that evolution drives perce... more The Bayesian approach of cognitive science largely takes the position that evolution drives perception to produce precepts that are veridical. However, some efforts utilizing evolutionary game theory simulations have shown that perception is more likely based on a fitness function, which promotes survival rather than promoting perceptual truth about the environment. Although these findings do not correspond well with the standard Bayesian approach to cognition, they may correspond with a behavioral functional contextual approach that is ontologically neutral (a-ontological). This approach, formalized through a post-Skinnerian account of behaviorism called relational frame theory (RFT), can, in fact, be shown to correspond well with an evolutionary fitness function, whereby contextual functions form that corresponds to a fitness function interface of the world. This fitness interface approach therefore may help provide a mathematical description for a functional contextual interface of phenomenological experience. Furthermore, this more broadly fits with a neurological active inference approach based on the free-energy principle (FEP) and more broadly with Lagrangian mechanics. These assumptions of how fitness beats truth (FBT) and FEP correspond to RFT are then discussed within a broader multidimensional and evolutionary framework called the extended evolutionary meta-model (EEMM) that has emerged out of the functional contextual behavioral science literature to incorporate principles of cognition, neurobiology, behaviorism, and evolution and are discussed in the context of a novel RFT framework called "Neurobiological and Natural Selection Relational Frame Theory" (N-frame). This framework mathematically connects RFT to FBT, FEP, and EEMM within a single framework that expands into dynamic graph networking. This is then discussed for its implications of empirical work at the non-ergodic processbased idiographic level as applied to individual and societal level dynamic modeling and clinical work. This discussion is framed within the context of individuals that are described as evolutionary adaptive and conscious

Acceptance and commitment therapy (ACT) has traditionally been employed at the individual level a... more Acceptance and commitment therapy (ACT) has traditionally been employed at the individual level as a therapeutic model to change the way the individual relates to their thoughts and feelings, helping to connect to meaningful values, and promoting psychological flexibility. More recently, it also has been applied at the societal level to promote prosocial behavior, facilitating shared values within organizations, and promoting a less coercive and more prosocial and equitable society. This has been done by applying evolutionary principles which can guide the selection of individual or group values that are adaptive and promote some advantage to the individual or group (e.g., promoting prosocial connection, compassion, and collective values), providing opportunities for multilevel impacts At an operational level, this work involves applying Nobel prize winner's

Background: There is much overlap among the symptomology of autistic spectrum disorders (ASDs), o... more Background: There is much overlap among the symptomology of autistic spectrum disorders (ASDs), obsessive compulsive disorders (OCDs), and alexithymia, which all typically involve impaired social interactions, repetitive impulsive behaviors, problems with communication, and mental health. Aim: This study aimed to identify direct and indirect associations among alexithymia, OCD, cardiac interoception, psychological inflexibility, and self-as-context, with the DV ASD and depression, while controlling for vagal related aging. Methodology: The data involved electrocardiogram (ECG) heart rate variability (HRV) and questionnaire data. In total, , participant's data of ECG recordings of healthy resting state HRV were recorded and grouped into age categories. In addition to this, another participants completed an online survey that included the following questionnaires: Yale-Brown Obsessive Compulsive Scale (Y-BOCS); Toronto Alexithymia Scale (TAS-); Acceptance and Action Questionnaire (AAQII); Depression, Anxiety, and Stress Scale (DAS); Multi-dimensional Assessment of Interoceptive Awareness Scale (MAIA); and the Self-as-Context Scale (SAC). Results: Heart rate variability was shown to decrease with age when controlling for BMI and gender. In the two SEMs produced, it was found that OCD and alexithymia were causally associated with autism and depression indirectly through psychological inflexibility, SAC, and ISen interoception. Conclusion: The results are discussed in relation to the limitations of the DSM with its categorical focus of protocols for syndromes and provide support for more flexible ideographic approaches in diagnosing and treating mental health Frontiers in Psychology frontiersin.org Edwards. /fpsyg.. and autism within the Extended Evolutionary Meta-Model (EEMM). Graph theory approaches are discussed in their capacity to depict the processes of change potentially even at the level of the relational frame.

Psychology has benefited from an enormous wealth of knowledge about processes of cognition in rel... more Psychology has benefited from an enormous wealth of knowledge about processes of cognition in relation to how the brain organizes information. Within the categorization literature, this behavior is often explained through theories of memory construction called exemplar theory and prototype theory which are typically based on similarity or rule functions as explanations of how categories emerge. Although these theories work well at modeling highly controlled stimuli in laboratory settings, they often perform less well outside of these settings, such as explaining the emergence of background knowledge processes. In order to explain background knowledge, we present a non-similaritybased post-Skinnerian theory of human language called Relational Frame Theory (RFT) which is rooted in a philosophical world view called functional contextualism (FC). This theory offers a very different interpretation of how categories emerge through the functions of behavior and through contextual cues, which may be of some benefit to existing categorization theories. Specifically, RFT may be able to offer a novel explanation of how background knowledge arises, and we provide some mathematical considerations in order to identify a formal model. Finally, we discuss much of this work within the broader context of general semantic knowledge and artificial intelligence research.

Public health (PH) messaging can have an enormous impact on shaping how individuals within societ... more Public health (PH) messaging can have an enormous impact on shaping how individuals within society behave, and can ensure it is in a safe and responsible way, consistent with up-to-date evidence-based PH guidelines. If done effectively, messaging can save lives and improve the health of those within society. However, unfortunately, those within Government PH bodies typically have little training about how to effectively represent PH messages in a way that is consistent with psychological theories of cognitive bias, in order to avoid cognitively biasing the public through their messages. As a result of this, inadequate representation of PH messages can result, which can often lead to cognitive bias in those from the public who read or listen to the message information. This can lead to poor decision making of the pubic as a whole, which can then further lead to harm and even death of public members as a result of these poor decisions. One way to minimize the problem of bias in decision making is to explore psychology theories that model how bias can occur from PH messaging, and identify ways in which PH agencies can utilize such approaches to improve the effectiveness of their messages. Previous focus has been largely on behavioral economic theories, however, here, other accounts are offered in addition to these. These include theories of heuristics and theories from the behavior analysis domain, which may increase the predictive power of modeling bias, and have applications for how best to represent PH message information which minimize bias.

Background: Alexithymia is a personality trait which is characterized by an inability to identify... more Background: Alexithymia is a personality trait which is characterized by an inability to identify and describe conscious emotions of oneself and others. Aim: The present study aimed to determine whether various measures of mental health, interoception, psychological flexibility, and self-as-context, predicted through linear associations alexithymia as an outcome. This also included relevant mediators and non-linear predictors identified for particular subgroups of participants through cluster analyses of an Artificial Neural Network (ANN) output. Methodology: Two hundred and thirty participants completed an online survey which included the following questionnaires: Toronto alexithymia scale; Acceptance and Action Questionnaire 2 (AQQII); Positive and Negative Affect Scale (PANAS-SF), Depression, Anxiety, and Stress Scale 21 (DAS21); Multidimensional Assessment of Interoceptive Awareness (MAIA); and the Self-as-Context (SAC) scale. A stepwise backwards linear regression and mediation analysis were performed, as well as a cluster analysis of the non-linear ANN upper hidden layer output. Results: Higher levels of alexithymia were associated with increased psychological inflexibility, lower positive affect scores, and lower interoception for the subscales of "not distracting" and "attention regulation." SAC mediated the relation between emotional regulation and total alexithymia. The ANNs accounted for more of the variance than the linear regressions, and were able to identify complex and varied patterns within the participant subgroupings. Conclusion: The findings were discussed within the context of developing a SAC processed-based therapeutic model for alexithymia, where it is suggested that alexithymia is a complex and multi-faceted condition, which requires a similarly complex, and process-based approach to accurately diagnose and treat this condition.

Background: Interoception and heart rate variability have been found to predict outcomes of menta... more Background: Interoception and heart rate variability have been found to predict outcomes of mental health and well-being. However, these have usually been investigated independently of one another.

Co-morbid mental health conditions such as anxiety, depression and fear avoidance are often assoc... more Co-morbid mental health conditions such as anxiety, depression and fear avoidance are often associated with chronic pain. This novel study aimed to explore the impact of osteopathic treatment on several psychological outcome measures relating to anxiety, depression, mental health and fear avoidance for a chronic pain population receiving osteopathic treatment over a 2-week period. The findings show that there were significant reductions in anxiety, pain, mental health dysfunction and improvements in self-care. These results are promising, and it is suggested that now a full-scale randomised controlled trial should be conducted.

Quarterly Journal of Experimental Psychology, 2015

We examined the effect of context on the learning of spatial coding in four experiments. Two part... more We examined the effect of context on the learning of spatial coding in four experiments. Two partially overlapping sets of stimuli, which had the very same stimulus–response spatial coding, were presented in unique contexts. Results show contextual locking—that is, response times to the very same item in a more common context (80%) were significantly shorter than those in a less common context (20%). Contextual locking was obtained both when the context was more salient (Experiments 1 and 2) and less salient (Experiments 3 and 4). In addition, results were obtained even when contextualization seemed less necessary (Experiments 2 and 4). Binding of information to context is discussed in relation to chunking, transfer effects, and practical applications pertaining to professional training.

Research in Developmental Disabilities, 2012

The ability to categorise items can be regarded as fundamental to many aspects of functioning, an... more The ability to categorise items can be regarded as fundamental to many aspects of functioning, and disruptions to this ability may have profound impacts on a person's ability to understand their environment (e.g., Bott, Brock, Brockdorff, Boucher, & Lamberts, 2006; Demetras, Post, & Snow, 1986). Thus, it is important to investigate Categorization abilities across a variety of populations, including in individuals with autism spectrum disorders (ASD). There are many forms of Categorization behaviour, which can be based on quite different mechanisms and principles, and it is important to explore each of these in the context of ASD. For example, some Categorization may be achieved on the basis of learning which items belong to which categories (supervised Categorization), and some categories are the result of intuitive groupings of items (unsupervised Categorization). To date, there have been no investigations of the latter form of Categorization for individuals with ASD, and the primary aim the current report is to document these abilities in individuals with ASD over a variety of circumstances. To make clear, it is important to study Categorisation in this context (i.e., with an ASD population) because there are several quite different explanations which try to explain the cognitive deficits behind ASD (e.g., such as Dube et al., 1999, for attentional deficits or Leader, Loughnane, Mc Moreland, & Reed, 2009, for retrieval deficits). For this reason, Categorization offers a useful means to experimentally assess the ability of this population to perform decisions which have not previously been explored (i.e., unsupervised decisions), thus this has real world applications. To this end, these methods could further our understanding of these cognitive deficits, and potentially lead to more refined screening methods. Supervised Categorization concerns learning pre-specified categories, where the experimenter determines which items belong to particular categories, and the participant must then learn to which category a new item belongs. In most cases, this is done by trial and error learning, with the participant receiving corrective feedback (e.g., Demetras, Post, & Snow, 1986). Reduced supervised Categorization ability has been noted in children with ASD (e.g., Klinger & Dawson, 2001). This is the difficulty in developing categories when feedback is given (i.e., supervised). In such a case, participants are given corrective

This article may be used for research, teaching, and private study purposes. Any substantial or s... more This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

The design of effective glyphs for visualisation involves a number of different visual encodings.... more The design of effective glyphs for visualisation involves a number of different visual encodings. Since spatial position is usually already specified in advance, we must rely on other visual channels to convey additional relationships for multivariate analysis. One such relationship is the apparent order present in the data. This paper presents two crowdsourcing empirical studies that focus on the perceptual evaluation of orderability for visual channels, namely Bertin's retinal variables. The first study investigates the perception of order in a sequence of elements encoded with different visual channels. We found evidence that certain visual channels are perceived as more ordered (for example, value) while others are perceived as less ordered (for example, hue) than the measured order present in the data. As a result, certain visual channels are more/less sensitive to disorder. The second study evaluates how visual orderability affects min and max judgements of elements in the...

Background: Myofascial release (MFR) is used to restore tissue extensibility of the fascia tissue... more Background: Myofascial release (MFR) is used to restore tissue extensibility of the fascia tissue and is considered to be useful in a number of clinical settings such as low back pain, ankle injuries, fibromy-algia, and headaches. There is, however, despite the popularity of MFR in manual therapy, little consensus on whether it leads to biomechanical, systemic or interoceptive outcomes. Aims: This study aimed to explore the immediate biomechanical (increased elasticity for increased range of motion), systemic (local vs. distal areas of pain threshold) and bodily awareness effects (interoception) of a myofascial release technique on the thoracic spine. Method: Twelve healthy participants took part in this triple-bind, repeated measures, cross-over design study, and were randomised into counterbalanced sequences of three conditions; a control, a sham, and the MFR condition. The outcome measures used were; range of motion (ROM), pain pressure thresholds (PPT), and interoceptive sensitivity (IS) to assess biomechanical, systemic, and interoceptive effects of MFR. Results: There were significant increases in ROM and PPT (both local and distal) post MFR intervention. There was also a positive correlation between baseline interoceptive sensitivity and post-MFR ROM and a negative correlation for baseline interoceptive sensitivity and post-MRF PPT. Interoceptive sensitivity did increase post-MFR but this was non-significant. Conclusions: The increase in ROM suggests that the MFR may have caused a biomechanical change in tissue elasticity creating an increase in tissue flexibility. The increase in both local and distal sites of the PPT suggest an overall systemic response to the therapy. The correlation between baseline IS and post-MFR ROM and PPT suggest that IS may be usefully applied as a predictor for ROM and PPT post-MFR.

Perspective-taking interventions have been shown to improve attitudes toward social outgroups. In... more Perspective-taking interventions have been shown to improve attitudes toward social outgroups. In contrast, similar interventions have produced opposite effects (i.e., enhanced negativity) in the context of attitudes toward elderly groups. The current study investigated whether a brief perspective-taking intervention enhanced with mindfulness would be associated with less negativity than perspectivetaking alone. One hundred five participants were randomly assigned to 1 of 4 conditions which comprised of an active or control perspective-taking component and an active or control mindfulness component. Participants were then administered an Implicit Associated Test to assess implicit biases toward the elderly. Results supported previous findings in that the condition in which perspective-taking was active but mindfulness was inactive was associated with greater negative implicit bias toward the elderly; however, some of this negativity decreased in the active perspective-taking and active mindfulness condition. The current findings and other mixed effects that have emerged from perspective-taking interventions are discussed from a Relational Frame Theory perspective.

BMJ Open, 2024

Quarterly Journal of Experimental Psychology, 2015

Research in Developmental Disabilities, 2012