Design and evaluation of a global workspace agent embodied in a realistic multimodal environment.

As the apparent intelligence of artificial neural networks (ANNs) advances, they are increasingly likened to the functional networks and information processing capabilities of the human brain. Such comparisons have typically focused on particular modalities, such as vision or language. The next frontier is to use the latest advances in ANNs to design and investigate scalable models of higher-level cognitive processes, such as conscious information access, which have historically lacked concrete and specific hypotheses for scientific evaluation. In this work, we propose and then empirically assess an embodied agent with a structure based on global workspace theory (GWT) as specified in the recently proposed "indicator properties" of consciousness. In contrast to prior works on GWT which utilized single modalities, our agent is trained to navigate 3D environments based on realistic audiovisual inputs. We find that the global workspace architecture performs better and more robustly at smaller working memory sizes, as compared to a standard recurrent architecture. Beyond performance, we perform a series of analyses on the learned representations of our architecture and share findings that point to task complexity and regularization being essential for feature learning and the development of meaningful attentional patterns within the workspace.

The influence of feature-based attention and response requirements on ERP correlates of auditory awareness.

In search for the neural correlates of consciousness (NCCs), it is important to isolate the true NCCs from their prerequisites, consequences, and co-occurring processes. To date, little is known about how attention affects the event-related potential (ERP) correlates of auditory awareness and there is contradictory evidence on whether one of them, the late positivity (LP), is affected by response requirements. By implementing a GO-NOGO design with target and nontarget stimuli, we controlled for feature-based attention and response requirements in the same experiment, while participants rated their awareness using a perceptual awareness scale. The results showed a prolonged auditory awareness negativity (AAN) for aware trials, which was influenced neither by attention nor by response requirement. The LP was affected by both attention and response requirements. Consistent with the levels of processing hypothesis, the LP was related to consciousness as a correlate of the processing of higher-level stimulus features, likely requiring access to a "global workspace." Our findings further suggest that AAN is a proper ERP correlate of auditory consciousness and thus a true NCC in the auditory modality.

A synergistic workspace for human consciousness revealed by Integrated Information Decomposition.

How is the information-processing architecture of the human brain organised, and how does its organisation support consciousness? Here, we combine network science and a rigorous information-theoretic notion of synergy to delineate a 'synergistic global workspace', comprising gateway regions that gather synergistic information from specialised modules across the human brain. This information is then integrated within the workspace and widely distributed via broadcaster regions. Through functional MRI analysis, we show that gateway regions of the synergistic workspace correspond to the human brain's default mode network, whereas broadcasters coincide with the executive control network. We find that loss of consciousness due to general anaesthesia or disorders of consciousness corresponds to diminished ability of the synergistic workspace to integrate information, which is restored upon recovery. Thus, loss of consciousness coincides with a breakdown of information integration within the synergistic workspace of the human brain. This work contributes to conceptual and empirical reconciliation between two prominent scientific theories of consciousness, the Global Neuronal Workspace and Integrated Information Theory, while also advancing our understanding of how the human brain supports consciousness through the synergistic integration of information.

The human brain consists of billions of neurons which process sensory inputs, such as sight and sound, and combines them with information already stored in the brain. This integration of information guides our decisions, thoughts, and movements, and is hypothesized to be integral to consciousness. However, it is poorly understood how the brain regions responsible for processing this integration are organized in the brain. To investigate this question, Luppi et al. employed a mathematical framework called Partial Information Decomposition (PID) which can distinguish different types of information: redundancy (available from many regions) and synergy (which reflects genuine integration). The team applied the PID framework to the brain scans of 100 individuals. This allowed them to identify which brain regions combine information from across the brain (known as gateways), and which ones transmit it back to the rest of the brain (known as broadcasters). Next, Luppi et al. set out to find how these regions compared in unconscious and conscious individuals. To do this, they studied 15 healthy volunteers whose brains were scanned (using a technique called functional MRI) before, during, and after anaesthesia. This revealed that the brain integrated less information when unconscious, and that this reduction happens predominantly in gateway rather than broadcaster regions. The same effect was also observed in the brains of individuals who were permanently unconscious due to brain injuries. These findings provide a way of understanding how information is organised in the brain. They also suggest that loss of consciousness due to brain injuries and anaesthesia involve similar brain circuits. This means it may be possible to gain insights about disorders of consciousness from studying how people emerge from anaesthesia.

Toward a universal theory of consciousness.

While falsifiability has been broadly discussed as a desirable property of a theory of consciousness, in this paper, we introduce the meta-theoretic concept of "Universality" as an additional desirable property for a theory of consciousness. The concept of universality, often assumed in physics, posits that the fundamental laws of nature are consistent and apply equally everywhere in the universe and remain constant over time. This assumption is crucial in science, acting as a guiding principle for developing and testing theories. When applied to theories of consciousness, universality can be defined as the ability of a theory to determine whether any fully described dynamical system is conscious or non-conscious. Importantly, for a theory to be universal, the determinant of consciousness needs to be defined as an intrinsic property of a system as opposed to replying on the interpretation of the external observer. The importance of universality originates from the consideration that given that consciousness is a natural phenomenon, it could in principle manifest in any physical system that satisfies a certain set of conditions whether it is biological or non-biological. To date, apart from a few exceptions, most existing theories do not possess this property. Instead, they tend to make predictions as to the neural correlates of consciousness based on the interpretations of brain functions, which makes those theories only applicable to brain-centric systems. While current functionalist theories of consciousness tend to be heavily reliant on our interpretations of brain functions, we argue that functionalist theories could be converted to a universal theory by specifying mathematical formulations of the constituent concepts. While neurobiological and functionalist theories retain their utility in practice, we will eventually need a universal theory to fully explain why certain types of systems possess consciousness.

The Conscious Nematode: Exploring Hallmarks of Minimal Phenomenal Consciousness in Caenorhabditis Elegans.

While subcellular components of cognition and affectivity that involve the interaction between experience, environment, and physiology -such as learning, trauma, or emotion- are being identified, the physical mechanisms of phenomenal consciousness remain more elusive. We are interested in exploring whether ancient, simpler organisms such as nematodes have minimal consciousness. Is there something that feels like to be a worm? Or are worms blind machines? 'Simpler' models allow us to simultaneously extract data from multiple levels such as slow and fast neural dynamics, structural connectivity, molecular dynamics, behavior, decision making, etc., and thus, to test predictions of the current frameworks in dispute. In the present critical review, we summarize the current models of consciousness in order to reassess in light of the new evidence whether Caenorhabditis elegans, a nematode with a nervous system composed of 302 neurons, has minimal consciousness. We also suggest empirical paths to further advance consciousness research using C. elegans.

Mientras que los componentes subcelulares de fenómenos cognitivos y afectivos que involucran la interacción entre experiencia, ambiente y fisiología -tales como aprendizaje, trauma, o emociones- son identificados con cada vez mayor detalle, los mecanismos biofísicos de la consciencia fenoménica permanecen elusivos. Nos interesa explorar si organismos sencillos como los nemátodos presentan consciencia mínima. ¿Hay algo que se sienta como ser un gusano? ¿O acaso los gusanos son máquinas carentes de toda experiencia? Los modelos "sencillos" nos permiten extraer datos de múltiples niveles en simultáneo: dinámica neuronal rápida y lenta, conectividad estructural, dinámica molecular, conducta, toma de decisiones, etc., y así testear predicciones de las propuestas teóricas actuales en disputa. En esta revisión compendiamos los modelos actuales de consciencia para evaluar, considerando la evidencia reciente, si Caenorhabditis elegans, un nemátodo con un sistema nervioso de 302 neuronas, tiene consciencia mínima. Sugerimos además vías empíricas para desarrollar investigaciones en consciencia utilizando C. elegans.

Further evidence and theoretical framework for a subliminal sensory buffer store (SSBS).

We recently provided evidence that strongly masked stimuli are not erased or overwritten but are briefly stored in a subliminal sensory buffer store (SSBS), where information can accumulate through repetition and become consciously accessible. SSBS supports a direct prediction made by the global workspace theory of consciousness (GWT) and has implications on discussions about conscious overflow and the problem of the criterion. Here we show that the presentation sequence and the time from the target presentation to evaluation does not significantly impact perception. We suggest that selected information from this subliminal sensory buffer store is transferred into a type of supraliminal short-term memory that keeps stable representations for longer durations with full conscious access. We argue that the level of conscious access of memory storage has a greater impact on subsequent reportability than initial phenomenology and needs to be included more prominently in discussions on perception and consciousness.

What if consciousness is not an emergent property of the brain? Observational and empirical challenges to materialistic models.

The nature of consciousness is considered one of science's most perplexing and persistent mysteries. We all know the subjective experience of consciousness, but where does it arise? What is its purpose? What are its full capacities? The assumption within today's neuroscience is that all aspects of consciousness arise solely from interactions among neurons in the brain. However, the origin and mechanisms of qualia (i.e., subjective or phenomenological experience) are not understood. David Chalmers coined the term "the hard problem" to describe the difficulties in elucidating the origins of subjectivity from the point of view of reductive materialism. We propose that the hard problem arises because one or more assumptions within a materialistic worldview are either wrong or incomplete. If consciousness entails more than the activity of neurons, then we can contemplate new ways of thinking about the hard problem. This review examines phenomena that apparently contradict the notion that consciousness is exclusively dependent on brain activity, including phenomena where consciousness appears to extend beyond the physical brain and body in both space and time. The mechanisms underlying these "non-local" properties are vaguely suggestive of quantum entanglement in physics, but how such effects might manifest remains highly speculative. The existence of these non-local effects appears to support the proposal that post-materialistic models of consciousness may be required to break the conceptual impasse presented by the hard problem of consciousness.

Assessing tests of animal consciousness.

Which animals have conscious experiences? Many different, diverse and unrelated behaviors and cognitive capacities have been proposed as tests of the presence of consciousness in an animal. It is unclear which of these tests, if any, are valid. To remedy this problem, I develop a list consisting of eight desiderata which can be used to assess putative tests of animal consciousness. These desiderata are based either on detailed analogies between consciousness-linked human behavior and non-human behavior, on theories of consciousness or on methods from human consciousness science. If a test or set of tests satisfies more of these desiderata, passing it provides stronger evidence of consciousness. Moreover, one can design future tests of animal consciousness with the intention of satisfying these desiderata to ensure their evidential strength.

The functional contributions of consciousness.

The most widely endorsed philosophical and scientific theories of consciousness assume that it contributes a single functional capacity to an organism's information processing toolkit. However, conscious processes are a heterogeneous class of psychological phenomena supported by a variety of neurobiological mechanisms. This suggests a plurality of functional contributions of consciousness (FCCs), in the sense that conscious experience facilitates different functional capacities in different psychological domains. In this paper, I first develop a general methodological framework for isolating the psychological functions that are associated with conscious experience. I then use this to show that the leading accounts-Global Workspace Theories, Higher Order Thought Theory and Information Integration Theory-all fail to acknowledge this functional pluralism, which limits their applicability as theories of consciousness.

A theory of consciousness from a theoretical computer science perspective: Insights from the Conscious Turing Machine.

This paper examines consciousness from the perspective of theoretical computer science (TCS), a branch of mathematics concerned with understanding the underlying principles of computation and complexity, including the implications and surprising consequences of resource limitations. We propose a formal TCS model, the Conscious Turing Machine (CTM). The CTM is influenced by Alan Turing's simple yet powerful model of computation, the Turing machine (TM), and by the global workspace theory (GWT) of consciousness originated by cognitive neuroscientist Bernard Baars and further developed by him, Stanislas Dehaene, Jean-Pierre Changeux, George Mashour, and others. Phenomena generally associated with consciousness, such as blindsight, inattentional blindness, change blindness, dream creation, and free will, are considered. Explanations derived from the model draw confirmation from consistencies at a high level, well above the level of neurons, with the cognitive neuroscience literature.

Embodied Intelligence: Smooth Coping in the Learning Intelligent Decision Agent Cognitive Architecture.

Much of our everyday, embodied action comes in the form of smooth coping. Smooth coping is skillful action that has become habituated and ingrained, generally placing less stress on cognitive load than considered and deliberative thought and action. When performed with skill and expertise, walking, driving, skiing, musical performances, and short-order cooking are all examples of the phenomenon. Smooth coping is characterized by its rapidity and relative lack of reflection, both being hallmarks of automatization. Deliberative and reflective actions provide the contrast case. In Dreyfus' classic view, smooth coping is "mindless" absorption into action, being in the flow, and any reflective thought will only interrupt this flow. Building on the pragmatist account of Dewey, others, such as Sutton, Montero, and Gallagher, insist on the intelligent flexibility built into smooth coping, suggesting that it is not equivalent to automatization. We seek to answer two complementary challenges in this article. First, how might we model smooth coping in autonomous agents (natural or artificial) at fine granularity? Second, we use this model of smooth coping to show how we might implement smooth coping in artificial intelligent agents. We develop a conceptual model of smooth coping in LIDA (Learning Intelligent Decision Agent). LIDA is an embodied cognitive architecture implementing the global workspace theory of consciousness, among other psychological theories. LIDA's implementation of consciousness enables us to account for the phenomenology of smooth coping, something that few cognitive architectures would be able to do. Through the fine granular analysis of LIDA, we argue that smooth coping is a sequence of automatized actions intermittently interspersed with consciously mediated action selection, supplemented by dorsal stream processes. In other words, non-conscious, automatized actions (whether learned or innate) often require occasional bursts of conscious cognition to achieve the skillful and flexible adjustments of smooth coping. In addition, never-conscious dorsal stream information and associated sensorimotor processes provide further online adjustments during smooth coping. To achieve smooth coping in LIDA we introduce a new module to the LIDA cognitive architecture the Automatized Action Selection sub-module. Our complex model of smooth coping borrows notions of "embodied intelligence" from enactivism and augments these by allowing representations and more detailed mechanisms of conscious control. We explore several extended examples of smooth coping, starting from basic activities like walking and scaling up to more complex tasks like driving and short-order cooking.

Hallucinations: A Functional Network Model of How Sensory Representations Become Selected for Conscious Awareness in Schizophrenia.

Hallucinations are conscious perception-like experiences that are a common symptom of schizophrenia spectrum disorders (SSD). Current neuroscience evidence suggests several brain areas are involved in the generation of hallucinations including the sensory cortex, insula, putamen, and hippocampus. But how does activity in these regions give rise to aberrant conscious perceptions that seemingly invade ongoing conscious experience? Most existing models assume that sensory representations are sometimes spontaneously activated in the brain, and that these spontaneous activations somehow play a causal role in the generation of hallucinations. Yet, it remains unclear how these representations become selected for conscious processing. No existing theory of hallucinations has specified such a "selection mechanism." Global Workspace (GW) theorists argue that the brain's interconnected processors select relevant piece(s) of information for broadcasting to other brain processors, rendering the information accessible to consciousness; this process known as "ignition" is associated with synchronized activity across distributed cortical and subcortical brain regions. Yet, it remains unclear how certain information and representations become selected for conscious processing. While GW theorists maintain that attention plays an important role, they have not delineated a formal "selection mechanism." This paper specifies a selection mechanism based upon two central hypotheses: (1) a functional network called the "salience network" plays a critical role in selecting sensory representations for conscious broadcast to the GW in normal (healthy) perception; (2) sensory representations become abnormally selected for conscious broadcast to the GW (instead of being filtered out of consciousness) in individuals with SSD that experience hallucinations.

Explanatory profiles of models of consciousness - towards a systematic classification.

Models of consciousness aim to inspire new experimental protocols and aid interpretation of empirical evidence to reveal the structure of conscious experience. Nevertheless, no current model is univocally accepted on either theoretical or empirical grounds. Moreover, a straightforward comparison is difficult for conceptual reasons. In particular, we argue that different models explicitly or implicitly subscribe to different notions of what constitutes a satisfactory explanation, use different tools in their explanatory endeavours and even aim to explain very different phenomena. We thus present a framework to compare existing models in the field with respect to what we call their 'explanatory profiles'. We focus on the following minimal dimensions: mode of explanation, mechanisms of explanation and target of explanation. We also discuss the empirical consequences of the discussed discrepancies among models. This approach may eventually lead to identifying driving assumptions, theoretical commitments, experimental predictions and a better design of future testing experiments. Finally, our conclusion points to more integrative theoretical research, where axiomatic models may play a critical role in solving current theoretical and experimental contradictions.

How does hemispheric specialization contribute to human-defining cognition?

Uniquely human cognitive faculties arise from flexible interplay between specific local neural modules, with hemispheric asymmetries in functional specialization. Here, we discuss how these computational design principles provide a scaffold that enables some of the most advanced cognitive operations, such as semantic understanding of world structure, logical reasoning, and communication via language. We draw parallels to dual-processing theories of cognition by placing a focus on Kahneman's System 1 and System 2. We propose integration of these ideas with the global workspace theory to explain dynamic relay of information products between both systems. Deepening the current understanding of how neurocognitive asymmetry makes humans special can ignite the next wave of neuroscience-inspired artificial intelligence.

An ESR Framework for the Study of Consciousness.

I will argue that, in an interdisciplinary study of consciousness, epistemic structural realism (ESR) can offer a feasible philosophical background for the study of consciousness and its associated neurophysiological phenomena in neuroscience and cognitive science while also taking into account the mathematical structures involved in this type of research. Applying the ESR principles also to the study of the neurophysiological phenomena associated with free will (or rather conscious free choice) and with various alterations of consciousness (AOCs) generated by various pathologies such as epilepsy would add explanatory value to the matter. This interdisciplinary approach would be in tune with Quine's well known idea that philosophy is not simple conceptual analysis but is continuous with science and actually represents an abstract branch of the empirical research. The ESR could thus resonate with scientific models of consciousness such as the global neuronal workspace model (inspired by the global workspace theory-GWT) and the integrated information theory (IIT) model. While structural realism has already been employed in physics or biology, its application as a meta-theory contextualising and relating various scientific findings on consciousness is new indeed. Out of the two variants: ontic structural realism (OSR) and epistemic structural realism (ESR), the latter can be considered more suitable for the study of consciousness and its associated neurophysiological phenomena because it removes the pressure of the still unanswered 'What is consciousness?' ontological question and allows us to concentrate instead on the 'What can we know about consciousness?' epistemological question.

The Hard Problem is Mainly Hard Work.

Doerig et al. point out that there is now a great deal of evidence bearing directly on our understanding of consciousness. However, they argue that the multiplicity of theories suggest that we have a 'lack of stringent criteria specifying how empirical data constrains ToCs.'

The global workspace theory, the phenomenal concept strategy, and the distribution of consciousness.

Peter Carruthers argues that the global workspace theory implies there are no facts of the matter about animal consciousness. The argument is easily extended to other cognitive theories of consciousness, posing a general problem for consciousness studies. But the argument proves too much, for it also implies that there are no facts of the matter about human consciousness. A key assumption is that scientific theories of consciousness must explain away the explanatory gap. I criticize this assumption and point to an alternative strategy for defending scientific theories of consciousness, one that better reflects the ongoing scientific practice. I argue there are introspectable inferential connections from phenomenal concepts to functional concepts that scientists can use to individuate the global workspace in terms of capacities that animals and humans share.

Involvement of top-down networks in the perception of facial emotions: A magnetoencephalographic investigation.

Conscious perception of the emotional valence of faces has been proposed to involve top-down and bottom-up information processing. Yet, the underlying neuronal mechanisms of these two processes and the implementation of their cooperation is still unclear. According to the global workspace model, higher level cognitive processing of visual emotional stimuli relies on both bottom-up and top-down processing. Using masking stimuli in a visual backward masking paradigm with delays at the perceptual threshold, at which stimuli can only partly be detected, suggests that only top-down processing differs between correctly and incorrectly perceived stimuli, while bottom-up visual processing is not compromised and comparable for both conditions. Providing visual stimulation near the perceptual threshold in the backward masking paradigm thus enabled us to compare differences in top-down modulation of the visual information of correctly and incorrectly recognized facial emotions in 12 healthy individuals using magnetoencephalography (MEG). For correctly recognized facial emotions, we found a right-hemispheric fronto-parietal network oscillating in the high-beta and low-gamma band and exerting top-down control as determined by the causality measure of phase slope index (PSI). In contrast, incorrect recognition was associated with enhanced coupling in the gamma band between left frontal and right parietal regions. Our results indicate that the perception of emotional face stimuli relies on the right-hemispheric dominance of synchronized fronto-parietal gamma-band activity.

Sedation Modulates Frontotemporal Predictive Coding Circuits and the Double Surprise Acceleration Effect.

Two important theories in cognitive neuroscience are predictive coding (PC) and the global workspace (GW) theory. A key research task is to understand how these two theories relate to one another, and particularly, how the brain transitions from a predictive early state to the eventual engagement of a brain-scale state (the GW). To address this question, we present a source-localization of EEG responses evoked by the local-global task-an experimental paradigm that engages a predictive hierarchy, which encompasses the GW. The results of our source reconstruction suggest three phases of processing. The first phase involves the sensory (here auditory) regions of the superior temporal lobe and predicts sensory regularities over a short timeframe (as per the local effect). The third phase is brain-scale, involving inferior frontal, as well as inferior and superior parietal regions, consistent with a global neuronal workspace (GNW; as per the global effect). Crucially, our analysis suggests that there is an intermediate (second) phase, involving modulatory interactions between inferior frontal and superior temporal regions. Furthermore, sedation with propofol reduces modulatory interactions in the second phase. This selective effect is consistent with a PC explanation of sedation, with propofol acting on descending predictions of the precision of prediction errors; thereby constraining access to the GNW.