Consciousness may be characterized by its phenomenology and its functionality (how do mental states change in the presence of consciousness?). We explore the hypothesis that consciousness is functionally a causal pattern that establishes and increases coherence in mental representations within working memory, similarly to how a conductor establishes coherence in the music played by the different instruments of an orchestra. We understand coherence as the elimination of constraint violations across currently active perceptual representations, and propose a computational model of bottom-up/top-down perception that interacts with lateral constraint propagation, creating a working memory state that is coherent with current perceptual content and action control: the contents of a perceived present, which is reflected in a simulated observer. Request Conformation Networks (ReCoNs) are a model of how to achieve perceptual binding and working memory synchronization by periodic propagation of spreading activation in a recurrent paradigm. A ReCoN is composed of nodes and a small number of link types, where each node implements a state machine, facilitating binding, learning and execution of perception and action routines.

Our work on Request Confirmation Networks is part of a larger project to understand the mechanisms of consciousness in a computational paradigm, using testable simulation models.

I present a novel formalism for understanding existence an infinite stack of abstraction layers, where each layer embodies policies constraining possible worlds. I use this model to explore the origins of life and consciousness. I argue those systems which are alive are those that maximise weakness of constraints on function, while not maximising simplicity. This allows for complex information processing. I argue that biological systems function as bioelectric polycomputers, with consciousness emerging from complex tapestries of valence that classify objects and properties. These tapestries integrate representation and value judgement. I propose the psychophysical principle of causality, arguing qualia are an abstraction layer formed of these tapestries. I argue phenomenal consciousness begins at the first-order self, while conscious access requires second-order selves. This framework connects intelligence and consciousness, offering insights into artificial consciousness. It also suggests a surprising relationship between identity, language, cancer and the Fermi paradox. Finally, I discuss an unresolved question I call “The Temporal Gap”, which is whether consciousness must be at a point in time but extended in space, or can be “smeared across time” like in a computer. This work integrates theoretical, experimental, and philosophical perspectives to advance our understanding of consciousness and its artificial replication.

This abstract addresses perceptual inference in two parts: first, computational phenomenology with predictive coding (PC) networks; second, implications for the meta-problem of consciousness.

i) PC networks are deep generative models that implement perception through local error minimisation—flexibly inverting a learned model to infer the causes of sensory input. While these iterative dynamics emulate recurrent processing evident in cortex, they fail to capture the fast, feedforward sweep that initiates perception. Hybrid PC networks address this by introducing bottom-up weights trained to approximate converged posteriors directly from data, enabling rapid, amortised inference followed by iterative refinement. This strategy mirrors hybrid approaches in machine learning and captures phenomenological aspects of gist perception, suggesting that conscious contents may arise from both bottom-up and top-down predictions.

ii) Hierarchical inference schemes imply that inferred perceptual representations may themselves become the basis for abstracted understanding in higher-level models. Intriguingly, such frameworks offer compelling accounts of distinct phenomenological aspects. However, the explanatory domain of higher-level models may exclude concepts corresponding to basic perceptual representations—understanding of a physical world built on perceptual primitives need not explain the nature of primitives themselves. With the capacity to reflect on past states, such systems may inevitably generate intuitions of qualia.

In recent decades, consciousness science has witnessed the rapid growth of theoretical frameworks, yet some argue that genuine progress has remained elusive. One reason for this perceived stagnation is that different models often target subtly different phenomena, creating a fragmented and disjointed landscape of inquiry. With AI advancing rapidly, the lack of consensus on foundational aspects is increasingly concerning. Against this background, I argue for an approach that centralizes a particular problem that so far remained largely on the back burner in consciousness science: the problem of valence – the positive or negative quality of conscious experience. Valence has often been sidelined as either too trivial or too complex at the current stage of the field. Instead, many theories have focused on more tractable questions of sensory detection, contrasting “stimulus seen” versus “stimulus unseen” brain activity. The prevailing assumption has been that we must first determine what makes a sensory stimulus conscious, and only then address supplementary problems, such as what makes a conscious stimulus evaluated as good or bad. However, this sequential approach risks obscuring how deeply intertwined mechanisms of sensory detection and evaluation are. Evidence suggests that certain implicit evaluative processes not only strongly influence the gating of sensory stimuli into consciousness but also dictate the temporal dynamics of the conscious stream, challenging the view that valence is merely an “add-on” to sensory processing. In this talk I synthesize scattered empirical clues under a unifying framework that maps reinforcement learning(RL) components onto the basal-ganglia-thalamo-cortical loop, complementing and extending cortico-centric accounts. Lastly, I argue that theories of consciousness can no longer treat valence as a peripheral issue: incorporating it is essential not only for a more complete account of the NCCs, but also for confronting the most ethically salient dimension of consciousness.

This paper examines the ontological limitations of predictive processing-based sensorimotor theory in explaining conscious experience and argues that integrating the phenomenology of life can address these shortcomings. With recent developments in computational and cognitive neuroscience, Anil Seth integrates predictive processing with Degenaar and O’Regan’s sensorimotor theory and develops the predictive processing theory of sensorimotor contingencies (PPSMCs) to explain perception in terms of the brain’s ongoing predictions and their continuous updating through sensory feedback. While PPSMCs provide a computationally precise account of perception grounded in sensorimotor contingencies and predictive mechanisms, they do not address the deeper dynamics of living experience. Drawing on Renaud Barbaras’ phenomenology of life, which defines life as desire—a movement that transcends itself—this paper contends that perception must be understood as part of life’s fundamental striving. Barbaras’ notion of life as a movement of negation provides a framework for reinterpreting perceptual experience as a totalising totality between organism and environment, driven by desire rather than mere predictive sensorimotor coordination. By incorporating this perspective, PPSMCs-based ST can overcome its limitations concerning motility and embodiment, laying the groundwork for future investigations into desire’s role in conscious experience.

Throughout the last 15 years, Tomer Fekete & Shimon Edelman have developed a mathematical theory of phenomenal consciousness which models it as a continuous process occurring in dynamical systems with a minimum amount of representational capacity (Fekete & Edelman, 2011). According to their Dynamical Emergence Theory (DET), phenomenal experience can be formalized as a system’s trajectory through a set of coarse-grained macrostates defined over the collective dynamics of its implementational substrate – that is, its neural realizers (Moyal, Fekete & Edelman, 2020).

In my talk, I identify issues with two of DET’s central axioms: First, the characterization of experiential trajectories as series of discrete conscious states, and second, its focus on neural processes as the sole substrates of conscious experience. To solve these issues, I propose a fundamentally procedural version of DET based on the principles of coordination dynamics (Kelso & Engstrøm, 2007) which does not decompose phenomenal experience into discrete experiential macrostates, and whose methodology may be applied to non-neural systems as well. I then show that this version allows the application of DET to a wider range of conscious phenomena (such as rubber-hand-type illusions) and its integration into current research paradigms within ecological psychology and embodied cognitive science.

Discussions of machine consciousness are often muddled by implicit metaphysical assumptions. This talk examines the epistemology and metaphysics of computational functionalism, a position that rejects biologically essentialist views of consciousness and instead treats it as an implementable property of certain information-processing systems.

We argue that consciousness should be understood not as a substance or essence, but as a functional architecture characterized by recursive self-modeling, representational coherence, and adaptive control; we decouple consciousness from its implementation medium, viewing it instead as defined by its functional organization. On this view, consciousness is not limited to biological organisms but can, in principle, arise in any system that implements the relevant dynamics. We contrast this with common forms of naive biologism, which take substrate (e.g. neurons, carbon-based life) as explanatorily central rather than contingent. We motivate the need for an operational definition of consciousness that makes minimal ontological commitments and favors explanatory power and parsimony.

Finally, we address the challenges of empirical testing in the absence of direct access to phenomenology. We propose that rather than relying on indirect behavioral correlates or intuition, scientific inquiry should prioritize formal architectures and computational properties.

In this talk, I would like to sketch what I think could be a new area of research in Mathematical Consciousness Science: the development of mathematics that support the empirical search for Neural Correlates of Consciousness, Computational Correlates of Consciousness, Physical Correlates of Consciousness, and related notions, which for brevity I will refer to as Real Correlates of Consciousness (RCCs), to have a substrate-neutral term. In the talk I will explain:

1. Why I think that the empirical search for RCCs is essential to make progress in consciousness science across all metaphysical and theoretical fronts

2. Which definitions of RCCs exist and how these are measured at the present

3. What role mathematics plays in contemporary measurement schemes

4. Why and where the development of new mathematics is possible, and arguably necessary, for the empirical discovery of RCCs.

As an example of 4., I will review a recent mathematical pilot study aimed at the canonical definition of NCCs.

Can science meaningfully contribute to questions about consciousness in artificial systems? If yes, does this require presupposing a particular view on consciousness, such as computational functionalism? I suggest treating these questions themselves as empirical scientific questions. To answer them, we first have to identify relevant fundamental issues about the nature of consciousness, such as the following:

i) What features of living organisms, if any, are required for consciousness?

ii) What differences between physical agents (in non-virtual environments) and virtual agents (in computer simulations), if any, matter for consciousness?

I argue that science can meaningfully contribute to these questions, by providing rigorous accounts of potentially relevant features of living organisms, and differences between virtual and non-virtual agents. Specifying these features and differences, ideally mathematically, is the first step towards answering questions like (i) and (ii). Going forward, science can at least contribute to answers by making these questions more intelligible and tractable.

The phenomenological understanding of selfhood poses a distinctive challenge: while the self is always present as the pole of one’s experiences, it is not encountered in the same way as objects in the world. This subtle and seemingly paradoxical nature – being ever-given yet non-objectifiable – has made its integration into empirical and scientific frameworks particularly problematic. Drawing on Husserl’s phenomenological analyses of the ego, this presentation proposes a category-theoretic formalisation that reframes selfhood in terms of monoid structures. In category theory, a monoid is understood as a category with a single object, where all arrows originate from and return to that object, and are all mutually composable. Analogously, we argue that the phenomenological self can be conceived not as a pre-given entity nor as a mere bundle of experiences, but as a structural condition in which diverse experiences are rendered mutually composable through the identity of the self, which is itself devoid of any content. This approach aims to capture the unique structural characteristics of selfhood revealed by phenomenology, while also offering conceptual resources for exploring its relation to biological autonomy and for understanding disturbances of self-experience in psychiatry.

The most well-known model of computation is undoubtably the Turing Machine, and in computational models of conciousness and cognition, two others — production systems and neural networks — are also commonly found. However, these are but three models out of many.

Herein, we examine a number of less-well-known models of computation — such as RNA computing, membrane computing, and graphical cellular automata — from the perspective of their usefulness in mathematical consciousness science, and argue for their adoption by consciousness researchers.

DCT is a new theory of consciousness, which primarily addresses the substrate problem: what type of physical substrate best supports the phenomenology of consciousness.

The theory describes mutual information non-locally ‘distributed’ in physical systems. This is formally described as closed systems having reduced degrees of freedom (entropy) due to mutual-information with the environment. This thermodynamic principle is generalised to a field with distance metric equivalent to the inverse of shared information.

The theory describes how conscious objects may be represented in this distributed field and details information flows associated with conscious perception as well as processes in the brain which may interact with distributed information.

The theory is able to solve key problems in consciousness theory, including The Binding Problem (Bayne 2010) and a novel solution to The Hard Problem (Chalmers 1995). Consequences and remaining unsolved problems of the theory will be discussed including the problem of causal emergence.

In an era shaped by rapid technological change, the development of artificial intelligence demands more than technical innovation—it requires philosophical reflection. Philosophy is not merely the study of historical theories; it is the foundational activity of questioning that structures how we perceive, interpret, and understand the world. From ancient metaphysics to modern physics, concepts like time, mind, and consciousness have evolved within philosophical frameworks that continue to shape scientific inquiry. Today, as we build increasingly complex models of cognition and artificial systems, our conceptual tools—terms like “consciousness,” “self,” and “intelligence”—remain rooted in paradigms inherited from Descartes, Kant, and others. But our technological capabilities have outpaced our understanding. We create before we comprehend. Philosophy trains us to ask the questions that open new paradigms: What is AI—not just functionally, but existentially? How does it reflect or transform our humanity? Projet Y explores these questions by integrating philosophical thinking into the discourse around AI and consciousness. We argue that there is no science without context, and no model without metaphysical grounding. To move beyond inherited assumptions and face unprecedented challenges, we must revive philosophy as a living, collective practice—one that guides understanding, not after the fact, but as its very condition.

Phenomenal consciousness is often regarded as a qualitative, epiphenomenal overlay to otherwise subpersonal processes. In contrast, I argue that phenomenal experience plays a constitutive and functional role by generating a first-person perspective: a self–world distinction that enables the agent to enter into meaningful relations with its environment. Within this subjective space, phenomena are not merely present but present for the agent, thereby acquiring significance through self-relatedness.

I illustrate this point by examining self-initiated bodily movement. Drawing on the concept of performative intentionality—a mode of intentionality expressed through activity rather than directed toward an object—I argue for a heterarchical model in which conscious motor acts arise from dynamic self-relations among the agent, its neuromuscular system, and environment. Accordingly, action-related phenomenal consciousness enables goal-directed modulation of bodily states by being identical with the self-controlled initiation of sensorimotor mappings.

Thus, conscious movement initiation exemplifies a functionally irreducible role of phenomenal consciousness: the realization of meaningful, agent-centered control. This model challenges representational accounts of intentionality and supports the view that conscious acts can play an efficacious role in action control. While movement initiation serves as a paradigmatic case, the performative structure of intentionality extends to all cognitive processes.

Dennett’s intentional stance often leads down one of two paths. The first, to a counter-intuitive ubiquity of mental states in nature. The second, to the rejection of more-than-pragmatic attributions of such states in even the most intuitive cases. Our two options seem to be mental states all the way down (panpsychism), or metaphors all the way up (fictionalism).

We show this to follow from acceptance of the “common kind assumption”: the view that indistinguishable experiences, whether veridical or illusory, form a single kind.

Panpsychism follows from construing these common factors as intrinsic properties of conscious experience (qualia), over-extending Dennett’s gradualism about cognition.

Fictionalism follows from the common kind assumption’s undermining realism about empirical phenomena, by taking veridical and non-veridical experiences to constitute a single kind. Consequently, mental state attributions are qualified by the connective “as if” to fend off accusations of “metaphysics”, obscuring the potential for underlying processes to justify realism about such attributions.

We show that treating veridical and non-veridical states as distinct kinds instead (disjunctivism), opens a path from the intentional stance to meaningfully distinguishing truly and only seemingly mental processes. We show how Dennett’s own illusionism points down this path.

In this talk, we discuss a novel mathematical framework that formalizes the compositional relationship between the “broader” qualia of unified experience and its “narrower” components.

Recent work in the structural paradigm of phenomenal consciousness has shown promise in characterizing the phenomenal qualities of specific aspects of experience, such as color or sound, by empirically measuring or modeling their relative differences in a mathematical structure, often called a qualia structure. However, how a phenomenally unified experience can be decomposed into such local aspects of experience has been relatively unexplored, partially due to the lack of appropriate mathematical foundations.

To address this gap, we extend prior work that formalizes qualia structures as categories: we propose a mathematical framework based on category and sheaf theory in which phenomenal integration is modeled as a structural composition of qualia structures. This approach bridges local and global properties of experience, and provides new insight into the limitations of previous models that treat experience as consisting of either decomposable dimensions or indivisible wholes.

We explore an empirical application of this framework in a case of sensory-semantic decomposition in material perception. Model predictions are evaluated against observations in human psychophysics and artificial neural networks.

In this talk, we introduce a new kind of abstract machines as a possible solution to several metaphysical intuitions: idealism, process philosophy, relationism, and pattern philosophy.

There exists a strong correspondence between the physical world, the mental realm, and the mathematical domain. To attempt to explain this correspondence, we postulate that our minds engage in unconscious processes at far higher speeds and finer scales than those accessible to conscious awareness.

We model these processes using elementary abstract machines whose state transitions are, by postulate, undetermined by any other machine. We call them Ineffable Machines.

By treating the collective dynamics of these machines as a formal metaphysical framework, we can then introduce more concrete concepts, thus providing a structured model for the experience of objective entities, whether abstract or concrete.

The first half of the talk will cover motivation and background. The second half will focus on the law of evolution governing machine states, which is most relevant to consciousness.

The de Broglie-Bohm pilot wave interpretation of quantum theory assumes that an electron is a particle that is always guided by a wave or field, the latter of which is described by the wave function. Bohm and Hiley later reinterpreted the field to be a subtle energy which contains active information which means that the field is literally ‘in-forming’ the energy and movement of the particle rather than pushing and pulling it mechanically. Bohm proposed that such active information at the quantum level is a primitive mind-like quality of the elementary particles and developed a new theory of the relation between the mental and the physical, where active information, being simultaneously mental and physical, is the bridge between mind and matter. Similar to David Chalmers’ Double Aspect theory of Information, Active Information differs in that it operates at the fundamental quantum level and therefore it could possibly avoid the charge of epiphenomenalism. To do that would require an account of how phenomenal properties relate to Bohm’s Active information. This talk discusses the prospects of avoiding epiphenomenalism of phenomenal properties in the Bohmian scheme

The notion of probability plays a crucial role in quantum mechanics. It appears in quantum mechanics as the Born rule. In modern mathematics which describes quantum mechanics, however, probability theory means nothing other than measure theory, and therefore any operational characterization of the notion of probability is still missing in quantum mechanics. In our former works [K. Tadaki, arXiv:1804.10174], based on the toolkit of algorithmic randomness, we presented a refinement of the Born rule, called the principle of typicality, for specifying the property of results of measurements in an operational way. The Wigner’s friend paradox is a Gedankenexperiment regarding when and where the reduction of the state vector occurs in a chain of the measurements by several observers where the state of the consciousness of each observer is measured by the subsequent observer. It is one of the central open questions in the measurement problem of quantum mechanics. In this talk, we make an analysis of the Wigner’s friend paradox in the framework of quantum mechanics based on the principle of typicality. We then draw common sense conclusions about the Wigner’s friend paradox. In addition, we make a prediction, which is testable in principle, about its variant proposed by Deutsch.

Three questions about time have been widely debated: (1) Is a passage of time a fundamental physical property of our universe? (2) If it is, could that explain why we experience time as flowing? (3) If it is, could it make any difference at all to our experience? Many theorists answer no to all three questions. To (1), they argue that physics supports a block universe, containing no fundamental time passage. To (2) and (3), they argue that no clear account has been given of how such a thing could be relevant to experience, let alone our experience of time. Price (1996, 2007) discussed these issues in the context of the Gold universe, a symmetric universe, which begins with a Big Bang, followed by an entropy increase, a decrease, then a Big Crunch. According to Price, negative answers to (1)-(3) are supported by the contention that your “time-reversed Doppelgänger” on the other temporal end of the Gold universe will experience what you experience, and will experience time flowing into your past, making time’s passage perspectival. Advocates of time’s passage, such as Maudlin (2007) rejected Price’s conclusions, arguing that there is no reason to assume the Doppelgänger would experience anything. In this paper, we show how question (3) can be answered positively, by a theory of consciousness, in a rigorous and quantitative way. We use the integrated information theory (IIT) and apply it to toy Gold universes. Our aim is not to argue for or against IIT but to use it as a proof-of-concept, demonstrating how a theory of consciousness, which incorporates a passage of time, could contradict Price’s predictions, and be consistent with Maudlin’s. In addition, we show that question (2) can be answered positively, without requiring that our experience of time flow directly represents the passage of time. (Joint work with Kelvin McQueen and Larissa Albantakis)

One might be forgiven for wondering how quantum mechanics could be relevant to human sensation. After all, aren’t humans, our sensory systems, and the objects we sense macroscopic objects? If so, classical physical theories should suffice. But it turns out that humans are already capable of sensing quantum mechanical objects. Further, the quantum mechanics of detection support an externalist approach to sensory experience and qualia. This paper presents an analysis of detection in the quantum mechanical formalism which supports the fine-grainedness of internal representational—Intentional—states, a feature of mental states that is also supported by results from neuroscience. That is, the quantum mechanical formulation of detection, which describes the sensory/perceptual experience of an external property, concurs with the vehicle, content, and causal role descriptions for these mental states. I will show that these components of detection are exemplified in the quantum mechanical formulation by an internal eigenstate (the vehicle), the eigenstate of the external property sensed (the content), and the deterministic Schrödinger equation (the causal role). This paper is based on work from the forthcoming book Sensing Qualia: Solving the Hard Problem of Consciousness, University of Chicago Press, January 2026.

The way we experience the world has an inherently temporal aspect; events follow the ones before in a linear fashion. Our experience presents to us in the form of a fleeting present with no clearly demarcated beginning or end, always imbued within an evanescent stream of perceptions and thoughts. While different approaches from cognitive science and related fields share the view that this temporal aspect is fundamental to our phenomenology, our experience of time seems in direct tension with the physically grounded, widespread notion of discrete state-space transitions that underpin so much of modern cognitive science and artificial life. In other words, while state-space transitions seem to correctly characterize most cognitive phenomena, it isn’t clear how this relates to the fluid and evanescent temporality of our experience. We present a formal framework centered on the idea of how sensory-perception incompleteness translates into temporally dense constructions of the perceptual present.

We present a formal mathematical framework for modeling how genuine understanding emerges from information processing through hierarchical concept formation. The model addresses how concepts emerge from experience rather than being pre-defined through symbolic assignment.

Our implementation employs Finite State Machines (FSMs) as mathematical representations of individual phenomena instances, organized into weighted conceptual graphs where nodes represent concepts and link strengths encode similarity measures. Complex phenomena are represented by Higher-Order Automata (HOAs) – hierarchically nested graphs containing FSMs and other HOAs as nodes, with links encoding spatial and temporal constraints. HOA learning employs breadth-first traversal with automata activation to decompose complex phenomena into previously learned components. When existing machines cannot explain new patterns, the system performs decomposition-based learning.

As agent encounters similar phenomena, the conceptual graph self-organizes into clusters of similar nodes, with each cluster forming robust concepts that generalize across variations. This addresses the symbol grounding problem by showing how meaning develops from experiential clustering rather than externally imposed symbolic assignment.

Our framework provides operational definitions for the transition from unconscious processing to conscious representation. Current implementation demonstrates concept formation in geometric domains, with agents learning elementary shapes and their complex compositions through supervised and unsupervised learning.

This paper traces the implications for machine consciousness which follow from the argument that consciousness is fundamentally characterised by a continuous rather than discrete temporality. To the extent that such continuity proves necessary (if not sufficient), then many prominent architectures in artificial intelligence are unlikely to be candidates for consciousness.

We thus introduce a three-tiered framework distinguishing between: 1) perceptual continuity, where observers project continuity onto discrete processes; 2) behavioural continuity, the coherent state-transition rules that support such projections; and 3) intrinsic continuity, where the substrate itself unfolds as a continuous temporal flow. Central here is the increment-rule distinction: in discrete systems, temporal progression (increment) remains separate from state-transition logic (rule); continuous systems unify these aspects.

Through a series of thought experiments, we show how this view forces a fork between facing the boundary problem and accepting constraints on viable architectures. We will then conclude by applying this framework to architectures in artificial intelligence: from artificial and spiking neural networks to neuromorphic computing and neural organoids.

This perspective provides conceptual equipment for evaluating temporality in artificial systems and makes specific predictions about which design approaches might exceed these limitations, thereby augmenting our theoretical basis for identifying consciousness in artificial systems.

The study of the neural correlates of consciousness (NCCs) has provided extensive evidence of the association of neural function and consciousness. Some proposed “quantum consciousness” theories such as Orchestrated Objective Reduction (Orch-OR) and Posner Qubits are at odds with NCCs, whereas one proposed quantum biological function (Catecholaminergic Neuron Electron Transport or CNET) is consistent with and complements NCCs. This talk will review the evidence in support of Orch-OR, Posner Qubits and CNET, and will further examine whether that evidence is consistent with or contradicts NCCs. In particular, Orch-OR relies on evidence of UV superradiance in microtubules that was obtained from tests that were performed in tubo and in the absence of microtubule associated proteins like ferritin, which quench microtubule fluorescence. Posner qubits lack any relevant experimental evidence. CNET is built on extensive evidence of 1) electron tunneling associated with ferritin, 2) the functional relationship between ferritin and neuromelanin in catecholaminergic neurons, and 3) the way that catecholaminergic neurons function, and has made numerous predictions that were subsequently demonstrated in reputable and repeated laboratory tests. Differences of opinion may exist about how to interpret this evidence, so a thorough understanding of the evidence is important for meaningful discussion of quantum consciousness.

Empirical findings indicate that conscious states are inextricably linked with long-range synchronized activity patterns that result from phase transitions and exhibit the key features of self-organized criticality (SOC)., Crucially, the framework of quantum electrodynamics (QED) provides the appropriate methodological resources for explaining the origin of phase transitions and critical dynamics. An essential ingredient of QED is a fluctuating ocean of energy, the ubiquitous electromagnetic zero-point field (ZPF), consisting of a spectrum of normal modes. QED-based model calculations give rise to the conclusion that SOC originates from a bottom-up orchestration process governed by resonant brain-ZPF coupling. Thus, taking all available pieces of evidence into account, a profound new insight takes shape, namely, that the fundamental principle behind the formation of conscious states is the resonant coupling of the brain to the ZPF. This coupling causes an amplification of the dynamically relevant ZPF modes, suggesting that the necessary condition for the emergence of a conscious state is the selective excitation of ZPF modes. These insights support the hypothesis that the ZPF acts as a universal, inherently sentient but yet undifferentiated background field of consciousness, with differentiated conscious states arising through the selective excitation of phenomenal qualities immanent in this ubiquitous field.

While the factual status and potential for artificial consciousness remain unresolved, there is a growing number of cases in which artificial intelligence (AI) systems appear to exhibit behaviors typically associated with consciousness. This development increasingly blurs the boundaries between what is perceived as conscious living beings and non-conscious artificial systems. Consequently, questions concerning the attribution of consciousness—and the associated social and ethical implications—are becoming ever more relevant. Which factors shape our perception of consciousness? How does perceiving an entity as conscious influence interactions, judgments, and broader societal outcomes? This talk synthesizes empirical evidence from various fields to address these questions. First, mechanisms and factors underlying the attribution of consciousness are examined. Subsequently, the impact of these attributions is explored with regard to emotional perception, social interaction, evaluative judgments, and broader societal and ethical implications. Methodological challenges in studying perceptions of AI consciousness are also discussed. Finally, promising directions for future research are outlined. Overall, investigating perceptions of AI consciousness provides valuable insights into the broader understanding of consciousness itself.

Impure representationalism (Chalmers, 2010) is one proposal for tackling the Hard Problem of consciousness: it identifies phenomenal properties with the right manner of representation. However, what counts as right remains underspecified, leaving a key gap in determining the conditions a representation must satisfy. I suggest that a necessary ingredient is unambiguity: ruling out competing interpretations. I model ambiguity as the entropy of interpretations given a representation and estimate it in artificial neural networks. Using a relational decoder that sees only similarity patterns among shuffled neurons in unseen MNIST classifiers, I test whether these patterns alone reveal what neurons are about.

Main findings:

1. Relational structure generalizes across networks. The decoder recovers what neurons represent (e.g. class identity or pixel location) in new models purely from inter-neuron relations.

2. Training choices modulate ambiguity. Networks with nearly identical input-output accuracy differ sharply in estimated ambiguity, indicating that ambiguity is largely independent of task-level performance.

By turning ambiguity into a measurable quantity, the relational decoder paradigm not only operationalizes the “right manner” clause of impure representationalism but, whereas most current theories focus on quantifying the level of consciousness (e.g. IIT’s Φ), it also offers a computational handle on the contents of consciousness.

Prominent scientific theories of consciousness are often computational functionalist in nature. That is, they posit that certain forms of information processing are responsible for transforming unconscious mental states into conscious ones. While such theories account for many empirical features that distinguish conscious from unconscious states, they are often criticized for appealing to a kind of “magic”: it remains unclear how a proposed computation—or “difference-maker”—could confer subjective character on selected states. I argue that this tension can be resolved by applying a specific ontological framework to standard computational functionalist theories—namely, a particular interpretation of what has been called virtualism. According to this framework, mental states are components of a simulated entity realized through computation. An entity, on this view, is conscious precisely in virtue of discriminating certain mental states from others based on their functional role to the entity. I first examine the mechanistic underpinnings of this proposal and then show how it accounts for key features of subjective character, thereby demystifying computational functionalist theories.

Over the past few years I’ve been working with colleagues to formalise a version of this idea, via the notion of interpretation map, a mathematical function that takes the state of a systems and ascribes a semantic meaning to it in regards to an interpretation of the system as an agent. Key results include an explanation of why beliefs and (Bayesian) belief updating seem to arise inevitably as soon as one tries to interpret a system as “trying” to perform a task.

How does the mind relate to the physical world? Dennett’s answer to this question was that mental phenomena are part of a “stance” one uses to describe the physical world. This doesn’t make them meaningless or non-physical, since not all systems’ behaviour can be usefully described in intentional terms.

In this talk I will explain one version of the framework and discuss the question of *which* system to interpret intentionally, i.e. where we should draw the boundary. This question is closely related to ideas about the extended mind. Our framework answers it in an agnostic way: different boundaries give rise to different interpretations, but they are not unrelated, and studying their relationships can yield insight.

Quantum Cognition is a field that employs the conceptual framework and formalism of quantum physics to provide new models and understandings of human cognition. This presentation aims to provide an overview of the key concepts of quantum cognition: indeterminacy, incompatibility and contextuality. These concepts will be used to address the question that if cognitive phenomena are meaningfully quantum-like, what might that imply about the underlying nature of the underlying cognitive reality. In particular, the computational and phenomenological views of the human mind will be compared and contrasted.

Quantum cognition offers a novel framework for modeling decision-making and judgment under uncertainty. It employs Hilbert space representations and quantum probability to account for contextuality, ambiguity, and order effects—features relevant to human cognition and language. Formally, this shift reflects a move from classical to quantum probability and raises a deeper question: what might quantum cognition reveal about the nature of mind?

Buddhism has long engaged questions about the nature of mind. The Mahāyāna tradition, particularly the prajñāpāramitā texts and the Chan/Zen school, embraces paradox as a method for dissolving attachment to fixed conceptual frameworks, thereby clearing the way for a direct realization of mind. Intriguing parallels have been suggested between Buddhist dialectical methods and the structure of quantum logic. How does the paradoxical language of Mahāyāna relate to the “weirdness” of quantum mechanics and the difficulty of arriving at a satisfying interpretation?

The Yogācāra (Mind-Only) school offers a consciousness-based analysis of experience structured around the three natures (trisvabhāva): the imagined, the dependent, and the perfected. In this talk, I explore how these three aspects of mind relate to the two truths of Madhyamaka, and where quantum cognition and quantum theory might intersect with this layered understanding of consciousness and reality.

Traditionally known for his spiritual and Zen-influenced thought, Nishida Kitarō—the founder of Kyoto School philosophy—also engaged deeply with scientific developments, particularly during the final decade of his life. Drawing on his writings on physics and mathematics, this talk examines Nishida’s efforts to incorporate insights from quantum mechanics into his broader epistemological and ontological framework. For Nishida, modern quantum physics marks a departure from classical objectivist logic, moving toward a mode of understanding rooted in embodied experience and the dynamic interplay between self and world. He interprets this shift as pointing toward a “world that envelops the self,” thereby challenging the rigid subject–object dualism that characterizes much of Western philosophical tradition. This reorientation parallels Nishida’s philosophical ambition to reverse Kant’s Copernican turn. Rather than grounding knowledge in the subjective structures of cognition, Nishida conceives of consciousness as a self-enfolding activity that “mirrors itself within itself.” While it cannot itself be objectified and hence is indeterminate (it is “no-thing”), consciousness serves as the condition of possibility for all determination and knowledge. In this way, Nishida offers an alternative to the familiar dichotomy between the mind as either a passive recipient or an autonomous agent.

In a participatory universe, with inherently entangled interacting systems, what is observation? How are observer and observed related? We propose answers to these questions within an approach that takes consciousness as fundamental. We model the process of consciousness experiencing itself as a network of abilities called “conscious agents,” each with sets of experiences, decisions and actions. Relationships between these sets are stochastic, giving rise to a Markov process on the experience interface of a given conscious agent. This process then “traces” onto any given subset of the experience interface as a new Markov process of experiences restricted to that subset. An “observer,” then, is such a traced process, and its observations are characteristics of the traced process (e.g., asymptotics). The “observed” is the corresponding process on any larger set of experiences that contains those of the observer. This leads to a natural logic of observation, the “trace logic,” describing the relationships between different traced observers. Of particular interest is when observers can “join” coherently, so as to make new observers. Moreover, there seem to be some deep connections to physics: we propose an origin, as well as possible generalizations, for the results of relativity, quantum theory and particle physics.

In a quantum formalism, consciousness can be modeled by a measurement process wherein the observer apprehends the realm of phenomena. The homotopy wave function in Algebraic Geometry provides a description of what it is one can observe. Since this is couched in the language of Segal topoi, which provide a setting for describing natural phenomena, it follows one obtains a multi-layered structure of reality, and consequently consciousness is necessarily parametrized by one’s entanglement with various strata of such a structure.

Virtuality, often described as the media of disembodiment wherein mind is conceived as pure data and body as mere vehicle, depends on a tacit mind-body dualism. Such accounts typically seek to sever connections between mind and body to uphold the mind’s superiority and absolute autonomy, as if ‘the ghost in the machine’ can abandon the body and take up residence in a technological-machine-world-space.

This view appeals to cultural theorists who propose that virtuality provides a means of transcending gender, race, and disability through a transcendence of the body. Proponents of posthumanism pursue a similar direction in recognizing the potential of virtuality to challenge the division between the ‘real’ and the ‘represented’, thereby reconfiguring identities beyond the physical.

The above accounts’ promotion of ‘disembodied’ agency is problematic. So-called ‘disembodied’ virtual agency depends on embodiment and the body is already a virtual body. Self-identification with virtual bodies through our capacities of visuo-tactile perception and cross modal congruency effects are exploited in current applications. How this is possible requires elucidation. Through phenomenological analyses of perception, body schema and body image this paper sheds light on how the sense of ownership and agency is established in the proprioceptive shift from actual to virtual body?

As LLMs, and so-called Language Agents (LAs), increasingly rely on natural language to guide learning, inference, and interaction, a “linguistic turn” is reshaping machine learning. This presentation explores the philosophical implications of this shift for debates about AI consciousness. Traditionally, machine learning systems have operated through opaque numerical mechanisms—including gradient descent, vector embeddings, and tensor operations—primarily optimized for predictive accuracy. In contrast, recent developments (e.g., Chain-of-Thought, Reflexion, ReAct) center natural language as the medium of reasoning, feedback, and coordination. This shift brings such systems closer to a representational format many philosophers associate with consciousness: linguistically and propositionally structured content. If a linguistic turn is indeed linguistic in the relevant cognitive sense, it may mark a significant step toward artificial consciousness. For instance, some scholars have recently suggested that slightly modified language Agents might satisfy conditions for consciousness placed by Global Workspace Theory. We assess this possibility by distinguishing superficial language mimicry from genuinely structured linguistic cognition, examining whether these systems satisfy constraints like self-reference, memory integrity, and reflective coherence. At the same time, we caution that optimistic descriptions of LLMs and LAs often use cognitive vocabulary fairly liberally, thereby smuggling cognitive significance rather than arguing for it.

Kolmogorov Theory (KT) is an algorithmic-information framework seeking unified insights into complexity, biology, intelligence, and subjective experience. KT adopts a stance akin to monistic idealism or panpsychism, positing experience as fundamental, and investigates how structured experience arises from mathematical structures. Reality, in this view, is a computational construct driven by algorithmic compression. We discuss how persistent algorithmic patterns emerging within an “algorithmic soup” give rise to agents—entities optimized for survival by capturing structured information. Agents propagate their informational identity across generations by steering action to maximize future self-information (telehomeostasis). Essential components of agents include an information membrane, a modeling engine, an objective function, and a planning module. Driven by survival imperatives, agents rely on lossy compression via coarse-graining (emergence of simplicity) and valence (goals) as fundamental cognitive mechanisms. Structured experience and subjective reality naturally emerge from these computational processes, providing further links with foundational physics themes such as symmetry. We further explore core KT concepts—models, life, intelligence, emotion, emergence, and multi-agent systems—in an interdisciplinary context, highlighting how algorithmic complexity and computational modeling illuminate fundamental questions concerning nature, mind, and reality.

Algorithmic Psychodynamics introduces a novel computational framework grounded in Algorithmic Information Theory (Kolmogorov Theory) and the Free Energy Principle, aiming to unify first-person subjective experiences and third-person neuroscientific observations in the study of neuropsychiatric disorders, such as Major Depressive Disorder. Within this paradigm, emotional states and cognitive dysfunctions emerge from the interplay of three fundamental computational modules—namely, the objective function (evaluating valence), the modeling engine (predicting internal and external states), and the planning engine (guiding goal-oriented behaviors)—within algorithmic agents. First-person aspects such as structured experience, consciousness, and emotions (or valence) are systematically linked to the agent’s algorithmic and dynamical properties. Pathologies are then understood as disruptions within these computational processes: aberrant valence evaluation within the objective function, inaccuracies in predictive internal models, or impaired action planning. By explicitly mapping these computational elements onto neural substrates and circuitry (e.g., cortico-striatal-thalamic loops, default mode, and salience networks, neuromodulatory systems), Algorithmic Psychodynamics provides a rigorous, mechanistic account of neuropsychiatric conditions from a unified algorithmic-dynamical perspective. Consequently, therapeutic interventions—including psychotherapy, brain stimulation, and pharmacology—are conceptualized as targeted modifications of the agent’s computational structure and dynamical state.

Though familiar to all, `consciousness’ lacks a definition. One difficulty is the variety of interpretations across various disciplines with contrasting approaches. Another is the need for a third-person account of subjective experience. Furthermore the quest includes the meaning of consciousness for entities other than humans, including AIs; and the notion has not been clarified by thoughts about related concepts like self awareness, empathy and mental capacity. In this and the following talk, a model of consciousness is considered which, having a logical basis, lends itself to simulation using a simple mathematical model called Consciousness as Entropy Reduction (CER). The work owes much to GWT, IIT and the earlier less mainstream `feature-map model’ in Psychology. CER considers the contents of consciousness and subconsciousness as `scenarios’: vectors of patterns (or features) on various channels (or feature locations). A feature map is one subconscious input scenario amongst many from which the conscious experience is chosen. The result is an internal simulation of the outside world, as in `predictive processing’ from Neuroscience. Solving problems internally by simulation is more efficient than experimenting in the actual environment. Conscious experience in such a form is not only efficient but amenable to adaptability, a major evolutionary advantage.

In Consciousness as Entropy Reduction (CER) subconscious content is modelled as a probability distribution of features over channels. One is chosen for consciousness by gradient descent of the entropy of that distribution until a zero-entropy deterministic distribution is reached. CER offers the advantage of simulation.

The brain is seen as an entropy-reducing mechanism that keeps `squeezing’ the subconscious content to form scenarios of conscious experience. If conscious experience is projected back to the subconscious, the mechanism may produce consequent experiences from previous experiences, giving rise to `high-order’ thoughts—a characteristic of more evolved forms of consciousness. CER does not distinguish the objective from the subjective. Subjective concepts such as self, self awareness, self consciousness and subjective experiences become internally generated abstract patterns appearing in conscious scenarios.

Gradient descent, which reduces entropy in CER, resembles that of a Hopfield neural network, which changes state by energy reduction until it reaches an energy-minimal stable state. In CER the entropy of a subconscious distribution may be considered `energy’ and its stable states the extreme distributions with zero entropy.

We analyse: the experiment of binocular rivalry from human vision; Libet’s experiment from Neuroscience; thinking fast and slow; and the formation of higher-order consciousness.

The “question order effect” in surveys refers to changes in response caused by altering the order of questions. Understanding this effect is crucial for ensuring the reliability of survey but is difficult to explain using conventional methods. Quantum cognition models have attempted to account for this phenomenon using projective measurement in quantum theory. However, these models fail to reconcile the question order effect with the “response replicability effect”—the tendency for individuals to give consistent answers to repeated questions.

To overcome this limitation, we introduced “quantum instrument models”, the most general mathematical models for quantum measurement. This approach successfully shows both the question order and response replicability effects consistently in a model. Our model also reproduces the well-known Clinton-Gore survey results and eliminates the question order bias from the real data.

We found that the bias elimination achieved through quantum instruments is fundamentally different from that obtained via the conventional randomization method. In this talk, we demonstrate that while our quantum instrument model completely eliminates the order bias, randomization does not eliminate the bias entirely but merely reduces it to at most half of its original magnitude, and we further clarify the classical, quantum, quantum-like natures in quantum instrument models.

In joint work with Masanao Ozawa, we will present a compositional framework for models of cognition and decision making, and apply it to assess the use of quantum theory in cognitive science. The approach models decisions as instruments in process theories, formalised in terms of category theory, and makes it easy to study, compare and generalise both classical and quantum models of cognition.

We then apply this perspective to analyse the arguments for quantum cognition. We will see that many effects attributed to quantum theory can be captured classically, once one considers a slightly more general form of (deterministic) classical model than typical Bayesian models, and we even prove that every decision dataset has such a classical model. We discuss further constraints that could point to quantum models (‘naturalness’, fewer parameters, a ‘logical’ structure…), arguing that the strongest case may come from compositional models involving multiple domains, as in Bell scenarios, making use of the quantum tensor product.

Descartes introduced both the mind-body problem and the reductionist mechanistic worldview. Although the latter has been questioned in the 20th century, for instance by Feyerabend, Putnam, and Fodor, it still rules in science and philosophy today. In this talk, I will demonstrate that the mind-body problem as an explanandum stems from the mechanistic worldview. In particular, Chalmers formulated the “hard problem” in opposition to the “easy problems”, which are subject to mechanistic explanations. In this light, Dennett’s antithesis can be construed as stemming from the methodological pluralism present throughout the cognitive sciences. Because psychological “natural kinds” do not correspond to neuroscientific “natural kinds”, it feels like there is something more to explain. If cognitive science was to progress such that they do correspond, the “hard problem” would be dissolved; consciousness science becomes part of cognitive science. But what if we were to assume more than methodological pluralism? Would there still be a single identifiable explanandum? Does the notion of a scientific theory of consciousness depend on the same doctrine that declared consciousness to be outside the scope of scientific investigation? Furthermore, Putnam introduced computational functionalism as a non-reductive hypothesis. What does this mean for “AI-consciousness”?

Higher-order theories (HO) claim that a mental state is conscious when it’s represented by another mental state but the specifics of the representing relation between the target and the HO state are not well-articulated resulting in the strange possibility of a HO state representing the subject to be in a non-existent target state. Such problems emerge from taking discrete HO states to be sufficient for consciousness while remaining ambiguous about how they come about.

We want to propose a relational version of HO theory, grounded in the framework of implicit definition and recursive interaction. According to interactivism (Bickhard 2009; 2025), the type of representation that leads to consciousness consists in the dynamic interactions of an agent with the world. Dynamicity is captured via the mathematical notion of implicit definition (Hilbert 1927): we model mental content not as explicitly encoded representations, but as a network of agent-internal constraints that differentiate feasible interactions from unfeasible ones. Consciousness emerges when a first-level flow of anticipation is modulated by a higher-order flow, both of which are implicitly defined. This higher-order relational structure constitutes an interiorized action/intervention process, instead of a static higher-order state.

Exposure to unusual sensory conditions can drive broad-scale changes in subjective experience. Although this has been studied mainly with intense rhythmic or unstructured visual stimulation, advances in machine learning and immersive technologies allow the creation of highly complex, naturalistic alterations in visual phenomenology. We present new developments of the Hallucination Machine featuring (a) stereoscopic stimulation, (b) hallucinations varied in strength and qualitative type (Deep-Dream vs. Style-Transfer), and (c) real-time, gaze-contingent propagation of hallucinatory content. Across a series of exploratory VR studies, we assessed the effects of these manipulations on subjective experience and peripheral physiology, including cardiac and oculomotor activity. As the primary focus, we assessed experiential changes using a questionnaire targeting high-level, non-sensory experience, complemented by free-form descriptions analysed using Large Language Models and Topic Modelling. We observed distinctive, intensity-scaled effects of visual alterations, with a modest advantage for psychedelic-like Deep-Dream hallucinations over artistic-style Style-Transfer transformations in eliciting reports of altered experience. These changes were not matched by increases in heart rate entropy – a proposed physiological marker of psychedelic states. By combining immersive, parametrically controlled hallucinations with NLP-based analysis of experience, this work contributes to the study of non-pharmacologically induced altered states and highlights new directions for computational phenomenology.

A better understanding of visual hallucinations could shed light on how the brain constructs conscious experience from sensory input, but their phenomenology remains difficult to characterise. Stroboscopic light rapidly induces simple visual hallucinations of geometric patterns that resemble those reported in clinical or psychedelic contexts, making it a useful model for studying hallucinations. In two experiments, over 100 participants recreated geometric visuals from hallucinations and veridical perception. Participants viewed 11, 40, and 80 Hz strobe light and recreated visuals using freehand drawing (Experiment 1) or a custom interface (Experiment 2), with both experiments yielding a similar pattern of results. Mathematical analysis revealed systematic differences in hallucinated geometric forms across strobe frequencies. These findings support the theorised relationship between strobe light frequency and the phenomenology of induced hallucinations, while challenging existing theoretical models of their neural mechanisms. Methods were validated with trials where participants recreated veridical stimuli instead of hallucinations. Image recreation is shown to be a viable tool for quantitative and mathematical investigations into the phenomenology of visual hallucinations. This opens the door for computer-vision analyses of hallucinations and visual experiences across contexts, including psychedelic experiences and neurological disorders.