Conference Agenda

Predictive processing theories propose that expectations generated from internal models shape visual perception. Consequently, individual differences in these models may account for the subjective ways we experience the world around us.

We used an inter-subject representational similarity analysis framework to explore whether idiosyncratic variation in scene perception and exploration can be predicted by similarities in individuals’ internal models of the world. To characterize these internal models, participants drew prototypical versions of indoor scenes. Inter-subject representational dissimilarity matrices (IS-RDMs) were constructed by quantifying similarities in these drawings across all pairs of participants, using feature maps extracted from a deep neural network.

Behavioral IS-RDMs were constructed capturing inter-subject similarities across various behavioral measures: (i) a categorization task involving indoor scene photographs, (ii) five subjective ratings (typicality, familiarity, attractiveness, usability, and complexity) of the same photographs, and (iii) gaze behavior during a free-viewing paradigm. Comparing IS-RDMs from the drawings with behavioral IS-RDMs revealed shared inter-subject similarities between participants’ drawings (and thus their internal models) and (i) categorization efficiency, as well as (ii) ratings of typicality, usability, and complexity. A preregistered replication confirmed the link between shared internal models and shared categorization performance. However, no significant correlation was observed between IS-RDMs derived from participants’ typical drawings and IS-RDMs based on inter-subject similarities in the fixation counts or gaze dwell times on objects.

Overall, we demonstrate a robust link between internal models and subjective variations in perception. Further research is needed to clarify the connections between internal models and scene exploration behavior.

Brain states, including attention and arousal, spontaneously fluctuate during wakefulness with implications on behavior and conscious perception. Pupillary dynamics track with arousal and attentional states because the brain regions regulating pupil size overlap with arousal and attention networks (e.g., brainstem and thalamus). Also, sensory neural processing is modulated by the arousal and attention systems. Therefore, pupil size can act as an accessible marker of brain activity linked with conscious and unconscious sensory processing. This study seeks to understand the relationship between pupillary fluctuations, sensory-evoked neural activity, and conscious perception. We aimed to (1) assess whether perception rate and behavior (e.g., reaction time) for visual and auditory stimuli correlate with pupil size trends (e.g., dilation versus constriction) and (2) examine if sensory-evoked brain activity – both perceived and non-perceived – is modified by spontaneous brain states indicated by pupil size trends. Healthy adult participants performed closed-loop visual and auditory detection tasks with simultaneous eye tracking, pupillometry, and magnetoencephalography (MEG). Task stimuli (images and tones) were presented when a custom real-time pupillometry software (rtPupilPhase) detected target pupil size trends. Behavioral results found that the phase of pupil size did not predict perception rate or reaction time. However, preliminary MEG analyses revealed that pupil phase corresponds with unique pre-stimulus activity across parietal and occipital scalp regions and differences in the amplitude and latency of sensory-evoked field potentials. These experimental findings support previous research and clarify how brain states linked to pupil size alter the neural processing of perceived and non-perceived sensory input.

Background:

Arousal fluctuations are ubiquitous, and substantially impact perceptual and cognitive performance. According to the Yerkes-Dodson law, optimal performance occurs at moderate arousal levels, forming an inverted U-shaped relationship. While we know that global arousal fluctuations affect task performance, humans also experience arousal cycles on shorter timescales. Recent evidence supports this law’s applicability to moment-to-moment arousal fluctuations across modalities (visual, auditory) and task types (discrimination, detection). However, the exact mechanisms how are unexplored. One key unknown factor is how stimulus strength shapes arousal effects on perception, with computational models suggesting stronger stimulus input (easier tasks) leading to optimal perceptual abilities with relatively higher arousal levels, compared to weaker input.

Methods:

This study investigates the effect of stimulus strength (task difficulty) on the arousal-performance relationship using an auditory discrimination paradigm with both easy and difficult variants (~90% and ~70% correct, respectively). We adjusted the target volume to each participant’s performance through staircasing, and monitored pupil size as an index of arousal. By applying hierarchical Bayesian probit regression modeling, we aim to characterize how the relationship between arousal and performance differs between input of different stimulus strength (easy vs. difficult tasks).

Results:

Preregistered hypothesis: we hypothesize that task difficulty modifies the shape and location of the arousal-performance relationship, expecting that task difficulty influences the curvature of this relationship, with increased curvature for the difficult compared to the easy task. Additionally, we expect the location of the point-of-peak-performance (pupil bin on x-axis) to differ between easy and difficult, e.g. rightwards for easy tasks.

Perception is often cast as an inferential process in which the brain combines noisy sensory input with predictions generated by internal models. One incarnation of this idea is predictive coding, in which the brain implements a hierarchy of generative models. Feedback connections carry predictions from higher level to lower level areas, and feedforward connections carry prediction errors. The brain would minimize these prediction errors by updating its internal models.

It has been proposed that the contents of conscious experience are determined by the current hypothesis about the state of the world that has the highest posterior probability. Therefore, the content of our conscious experience should be determined by the interactions between predictions and prediction errors.

We have designed a task in which expected and unexpected auditory stimuli were presented against a noisy background, rendering their detection challenging. We aimed to test the hypothesis that, in line with predictive coding, the detection of expected tones should be associated with low levels of prediction error, because sensory evidence matches with the prediction at hand. Instead, the detection of unexpected tones is only possible when they elicit enough prediction error to successfully update the mistaken prediction.

By acquiring magnetoencephalography (MEG) and behavioral data during the execution of this auditory task, we can relate the amplitude of the mismatch response to the detectability of expected and unexpected tones. With this study, we hope to clarify how predictions and predictions errors interact to determine the conscious perception of sounds.

Predictive processing has emerged as a dominant framework for understanding consciousness and cognition. Predictive processing has been shown to be highly distributed within and across brain areas in humans and primates (10.1038/s41467-024-48329-7) but the question remains as to whether a similar process occurs for simpler, more phylogenetically distant animals. In this study, we investigated the distribution of prediction errors across the brain of Drosophila melanogaster. Using electrophysiological recordings from 16 brain regions in response to expected and unexpected visual stimuli, the distribution of prediction errors was calculated using Co-Information. This information-theoretic measure calculates how much prediction error information is shared between areas (redundancy), and how much is only available as a combination of two areas’ activity (synergy). Our results revealed patterns of Drosophila synergistic and redundant prediction error signals strikingly resembling those observed in common marmosets. In both cases, synergistic and redundant interactions were present within as well as across different stages of the processing hierarchy. Synergy, however, was mainly present between early and late time-points, with respect to stimuli presentation, whereas redundancy was stronger for more proximal time-points. This result suggests a fundamental property of predictive information processing, present across brains of vastly different sizes and structure. Evolutionarily, this might reflect the importance of robustness to noise, as well as that of synergistically binding features of stimulus and prediction details together. Moreover, if predictive processing is a key feature of the development of consciousness, this result might have implications for the breadth of consciousness across the animal kingdom.

False perceptions, such as hallucinatory-like experiences, provide a compelling framework for exploring the mechanisms underlying perceptual and metacognitive processes. We investigated how prior expectations and feedback on task performance influence false detections of faces and their associated confidence.

Participants performed a face detection task in which faces were embedded in visual noise, followed by a confidence rating. Prior expectations were manipulated using a cue at the beginning of each trial, indicating the likely presence or absence of a face with 80% validity. Two task variations were used: one without feedback and one with trial-by-trial feedback (correct vs. incorrect).

In both tasks, participants made false alarms (10% with feedback and 7% without feedback). False alarms in the no-feedback condition were associated with higher confidence, with 57% rated as medium or high confidence, compared to only 32% in the feedback condition. Priors influenced false alarms only in the feedback task, where they modulated both detection rates and confidence ratings. We then tested how Bayesian models of detection and confidence, incorporating different feedback-prior integration mechanisms, explain our detection and confidence results.

These preliminary findings suggest that feedback and priors jointly shape both perceptual and metacognitive judgments.

Arousal plays a crucial role in shaping conscious perception, yet its influence on the neural mechanisms underlying this process remains underexplored, particularly in humans. Recent research has shown that pupil-linked arousal modulates the neural representation of visual information, but it is unclear whether this modulation differentially affects feedforward and recurrent processing. Given the proposed role of recurrent activity in conscious perception, understanding how arousal shapes these distinct neural dynamics is essential. Behaviorally, arousal effects often follow an inverted U-shaped pattern, as described by the Yerkes-Dodson law, but how this translates to neural processes remains poorly understood. This preregistered study aims to bridge this gap by examining how pupil-linked arousal influences feedforward and recurrent neural processes in humans, using EEG and pupillometry.

Participants performed a detection and discrimination task with masking, requiring either detection of the Kanizsa illusion or discrimination of local image contrast. Pre-stimulus pupil size is used as an index of arousal, which is then linked to the decoding of visual features — local contrast (feedforward) and the Kanizsa illusion (recurrent). Polynomial regression models are applied to evaluate the relationship between arousal and these neural processes.

Our preregistered hypotheses are that feedforward processing will remain stable across arousal levels, while recurrent processing is expected to exhibit an inverted U-shaped relationship with arousal, reflecting the Yerkes-Dodson principle. These findings aim to provide new insights into the interplay between arousal and neural dynamics, offering a more detailed understanding of the mechanisms underlying arousal-related fluctuations in conscious perception.

Hybrid predictive coding (HPC) extends predictive coding (PC) networks with an extra set of bottom-up weights connecting the same hierarchical node layers. By treating converged activations (variational posteriors) as targets of a supervised learning task, these weights enable amortised inference, learning to map data directly onto approximate posteriors in a feedforward pass. While PC provides an appealing Helmholtzian model of perception as inference, the iterative prediction error minimisation (PEM) dynamics with which it attains variational posteriors are slow and computationally expensive; in contrast, humans can reliably perform certain visual tasks at extremely fast timescales. Analogous to the neurophysiological feedforward sweep, HPC networks can trigger a feedforward pass upon detection of stimulus change to gain a rapid gist of the scene, priming subsequent iterative inference. Importantly, the posteriors of an amortised scheme are qualitatively distinct from evolving variational posteriors. We reasoned that in the context of gist perception, amortised inferences would make systematic and distinctive errors for improbable or incongruent features of scenes. We show how real-time decoded activations and generative reconstructions of HPC network activity successfully model existing results in the gist perception literature, and we introduce a more biologically plausible ‘Natural HPC’ variant for modelling data from ongoing and planned psychophysical experiments. This work—under the banner of computational neurophenomenology—supports the wider hypothesis that conscious contents may be described by posterior distributions driven by predictions flowing in bottom-up as well as top-down directions, presenting new targets for the identification of predictive processing mechanisms in the brain.

Theoretical accounts postulate that the catecholaminergic neuromodulator noradrenaline shapes cognitive behavior by reducing the impact of prior expectations on learning, inference, and decision-making. A ubiquitous effect of dynamic priors on consciousness perception under uncertainty is choice history bias: the tendency to systematically repeat, or alternate, previous choices, even when stimulus categories are presented in a random sequence. Here, we directly test for a causal impact of catecholamines on these priors. We pharmacologically elevated catecholamine levels through the application of the noradrenaline reuptake inhibitor atomoxetine. We quantified the resulting changes in observers’ history biases in a visual perceptual decision task. Choice history biases in this task were highly idiosyncratic, tending toward choice repetition or alternation in different individuals. Atomoxetine decreased these biases (toward either repetition or alternation) compared to placebo. Behavioral modeling indicates that this bias reduction was due to a reduced bias in the accumulation of sensory evidence, rather than of the starting point of the accumulation process. Atomoxetine had no significant effect on other behavioral measures tested, including response time and choice accuracy. We conclude that catecholamines reduce the impact of a specific form of prior on consciousness perception.

Eye tracking (ET) has become an essential tool in cognitive research, allowing insights into perception, attention, and behavior. A critical step in ET analysis is parsing raw data into discrete eye movements (EMs) using detection algorithms (detectors). These detectors differ significantly in their ability to identify EM types and their overall performance, necessitating a standardized framework for evaluation.

Our work introduces pEYES, an open-source Python toolkit designed to streamline ET data analysis and EM detection, as well as provide robust, reproducible comparisons between detectors. pEYES integrates widely used evaluation metrics and visualizations, offering a simple programming interface for both novice and advanced ET researchers.

To demonstrate pEYES’ capabilities, we evaluated seven commonly used threshold-based detection algorithms using a publicly available human-annotated dataset of free-viewing trials. Evaluation criteria included label agreement with the human-annotated ground truth (Cohen’s Kappa), temporal alignment with the annotated data (relative timing offset and deviation), and the ability to distinguish fixation and saccade onsets and offsets (discriminability index, d').

The results revealed significant variations in detector performance. Engbert’s algorithm, leveraging an adaptive velocity threshold, outperformed other detectors in fixation onset and offset detection, achieving precision comparable to human annotators. It showed slightly lower performance for saccades, though still competitive. Conversely, spatial dispersion-based detectors, like I-DT, consistently underperformed across metrics.

This study underscores the importance of selecting task-specific detection algorithms and highlights the need for transparent, reproducible evaluation pipelines. pEYES represents a versatile and freely available resource for advancing ET research and enhancing detector performance evaluations.

Humans and animals are able to rapidly and accurately estimate the number of objects or individuals presented in a visual scene. While numerosity perception has been widely studied using simple well controlled inanimate stimuli such as dots, little is known regarding numerosity processing of social animated agents, which detection is facilitated by a specialized brain system.

Extending visual numerosity paradigms from dots to the social domain, we recently designed a numerosity estimation task for human stimuli (NEH) in immersive virtual reality (VR), revealing classical behavioral numerosity estimation effects in this NEH (i.e., perceived numerosity increases with cardinality, overestimation above the subitizing range), and showed that the robotically induced mental state of a presence hallucination (PH) further increases NEH overestimation.

Here, by combining our VR-robotic platform and NEH task with electroencephalography, we investigated the neural correlates underlying the increased NEH overestimation associated with robot-induced PH. Our visual-evoked potential data showed that robot-induced PH are associated with an increased early P2p component, dominating over left posterior electrodes, which has typically been associated with numerosity estimation of dots. Furthermore, the P2p amplitude correlated with NEH overestimation. Finally, source reconstruction localized the P2p modulation within the left extrastriate body area, a specialized brain region for human body perception. These results suggest that NEH is processed in pathways for social stimuli and propose a novel neural marker indexing robot-induced PH.

When searching for an object we must mentally represent its features as a search template, however, many objects often contain multiple identifying features from a range of dimensions and with different levels of typicality/probability. To test whether participants represent multifaceted objects as their most typical and salient features, participants had to search for a target category consisting of multiple shapes (dresses, shirts) and colours (purple, green), which varied in their frequency. One feature within each dimension was high probability (2/3 of trials) and one low probability (1/3 of trials). To probe which features of the target were prioritised, the isolated features were presented as task-irrelevant visual probes, which were either spatially valid or invalid cues to target location. Behavioural evidence revealed that all partially matching visual probes influenced spatial attention, resulting in faster target detection when spatially congruent. Electrophysiological data also revealed that all probe stimuli were processed, though at different stages. Specifically, all probes elicited a lateralised positive component (350–500ms latency range), consistent with the Late Directing Attention Positivity (LDAP) that reflects the cued strategic deployment of attention to the target. However, only visual probes congruent with the target shape elicited an earlier N2pc component (250–350ms latency range), consistent with attentional capture prior to strategic allocation. Interestingly, target feature probability had very little influence on the magnitude of attentional processing at any level, suggesting that when possible we appear to represent all target relevant features to detect a complex target, rather than tuning towards a singular high probability feature.

Background: Consciousness enables humans to flexibly apply sequential strategies to solve hierarchical problems (Kahneman, 2011). However, the origins of these reasoning capabilities remain unclear: Can sequential strategies emerge purely through training on hierarchical tasks, or do they require explicit architectural biases? Here, we examine whether recurrent neural networks (RNNs) trained via backpropagation develop sequential processing strategies in a hierarchical task without explicit architectural constraints.

Methods: We train small vanilla RNNs on visual counting tasks. Stimuli are binary patches appearing on a spatial grid, categorized as targets or distractors based on a fixed Boolean function. The RNN must report the number of targets at trial end. For comparison, we introduce various biases toward sequentiality by presenting stimuli in sequence or requiring sequential classification.

Results: In the static counting condition, backpropagation strongly favors parallel internal representations, mapping stimuli onto separate, minimally interacting hidden neuron clusters. As the stimulus count increases, this parallel approach exceeds capacity, impairing performance. In contrast, sequential input or output conditions promote sequential computational strategies, facilitating learning.

Conclusions: Our preliminary results suggest that backpropagation in RNNs fails to discover sequential solutions for static hierarchical problems. Future work should explore how this effect scales with network size and stimulus count and whether multiplicative interactions enable sequential solutions by discovering attention (Zucchet et al., 2023). Finally, in the absence of sequential task conditions, explicit reasoning systems for hierarchical tree search may be necessary, with loose parallels to Large Reasoning Models (Besta et al., 2025) and implications for theories of top-down attention.

Time-frequency dynamics derived from hierarchical frequency tagging methods can be used to delineate the mechanisms underlying active inference processes [1,2]. This study asks if similar dynamics are directly associated with the emergence of conscious perception and crucially whether and how actions (key to active inference approaches to conscious content [3]) modulate these dynamics. By coupling EEG frequency tagging with the classic motion induced blindness paradigm (MIB), in which the stimulus remains consistent but only one’s conscious content is altered, we aim to disentangle these processes. We use a carefully controlled adaptation of MIB to investigate the unfolding of SSVEP and SWIFT frequency tags when a stimulus reappears into conscious awareness. In an ‘active’ condition participants were asked to look towards a target stimulus following its disappearance from consciousness. Replay trials emulating active trials are then used to create a ‘passive’ condition in which the eyes remain fixated at the centre of the screen while the stimulus is moved toward their eye. We find distinct differences between Active and Passive in the temporal fluctuation of our frequencies of interest, as well as their associated intermodulation frequencies data beyond those attributable to low-level states (arousal etc.). Our findings demystify the temporal dynamics of the reappearance of conscious content within the brain and support the use of active inference as an account of conscious content.

We consider the problem of uncovering the time course of neural activity leading to self-initiated movements, and of the degree to which such activity can be used to predict when a movement is imminent. We address a major pitfall common across decades of experimental research on self-initiated movement, namely that the data epochs subjected to analysis are only those culminating in a movement. We propose a rigorously controlled experimental paradigm that yields epochs that either do or do not terminate with a self-initiated movement but are well-matched in other respects. We applied this framework as a basis for experiments in which we recorded M/EEG data from human participants, and then used machine learning in a sliding window to classify data segments as belonging to a movement or non-movement epoch. When we tried to classify a window of data as belonging to a remote window of reference versus some other temporal offset closer to movement, as is currently standard in the field, the results suggested movement could be accurately predicted more than a full second before movement onset. By contrast, when we instead included no-movement epochs, this kind of extreme result was abolished; rather, we observed that classification accuracy rose abruptly from near chance to near ceiling as movement onset approached, and at time scales more in line with what participants report in such tasks. As these and further analyses show, our framework provides an improved vantage point for studying fundamental characteristics of neural activity specifically directed toward self-initiated movement.

Affective phenomena—such as sensation, emotion, and mood—are often considered telltale signs of consciousness. However, there is a dire lack of consensus in the study of affect itself. To address this, 173 researchers from 23 countries worked “Towards an Integrated Understanding of the Human Affectome” (Neuroscience and Biobehavioral Reviews). The resulting capstone integrates disparate perspectives into a common set of metaphysical and mechanistic assumptions about affective phenomena—via a principle regarding their intrinsic purpose. Thus, the Human Affectome framework enables the field to systematize affective phenomena, understood as algorithms at the level of the entire human agent. When subsequent commentaries called for clarity on the role of consciousness, we noted that phenomenology provides structure with which to map those algorithms, effectively tracking the contents of consciousness. However, the converse question remains: does affect help explain consciousness? While the ‘hard’ problem of consciousness is famously intractable, with the ‘easy’ problems only somewhat easier, we raise another challenge inherent to both: why do contents of consciousness relate to each other such that there is structure to experience at all? We consider this the ‘soft’ problem—not only because answers must be informed by “fluffy” subjective methods, but also because, at its heart, the issue concerns the texture of experience. In positing the agentivity of feelings as both their defining nature and organizing principle, the Human Affectome framework inadvertently uncovers this soft problem. Accordingly, affective phenomena may be especially convincing markers of consciousness precisely in virtue of being strong indicators of underlying agentive organization.

It has long been recognised in econometrics that certain classes of time series, specifically those exhibiting non-stationary, random walk-like behaviour (e.g., stock prices), will yield spurious results for statistics such as correlation, which are based on linear regression modelling. While regression analyses may be rescued by differencing the time series---thus enforcing stationarity---this comes at the cost of discarding potentially informative statistical structure in the data. One example of this is cointegration: a situation where variables in a multivariate system ``random walk together''. The analysis of cointegration relationships in neurophysiological data therefore offers a new avenue for discovering potentially novel functional relationships in the brain, which are inaccessible via conventional correlation-based analyses.

Cointegration analysis has been widely deployed in econometrics, but not yet in neuroscience. On the face of it, it might appear that cointegration analysis will be inappropriate for neurophysiological processes, since it presupposes random walk-like, and therefore unbounded behaviour, whereas biophysiological processes are invariably homeostatically constrained. Addressing this, we take the view that neurophysiological processes may exhibit cointegration-like behaviour over short time scales; that in the short run, neural activity at specific brain regions may ``random-walk in step''. We develop a principled statistical approach to quantifying short-run cointegration for bounded processes at increasing time scales, along with a geometric interpretation of cointegration relationships appropriate to neurophysiological dynamics. We illustrate our approach using resting-state MEG data.

Demonstrating unconscious processing is challenging, and the extent to which perceptual and cognitive functions can be executed unconsciously remains debated. Two requirements are needed for establishing unconscious effects: First, showing that the relevant stimulus or feature was processed; Second, providing evidence for the absence of conscious awareness of the stimulus/feature. The latter has been criticized due to low statistical power in awareness tests, raising concerns that reported unconscious effects may be contaminated by conscious processing. To improve the sensitivity to conscious contamination, we propose two awareness tests: the frequentist Group Binomial or Chi (GBC) test and the Bayesian Group Binomial Bayesian (GB-Bayes) test. These tests are designed to maximize power when applying objective awareness measures. Using simulations, we compared their sensitivity and specificity to commonly used methods: t-tests, Mixed Model Logistic Regression, and Bayesian t-tests. Our results demonstrate a clear power advantage for the GBC and GB-Bayes tests across different simulated scenarios. Additionally, we reanalyzed 79 previously reported unconscious effects from 15 studies, which further supported the power advantage of these tests. Our systematic evaluation of these tests provides a robust framework for assessing awareness tests in general to mitigate the risk of conscious contamination.

Experimental manipulations of conscious sensory perception ordinarily rely on behavioral metrics that can be used to infer the presence or absence of consciousness. Broadly, these can be grouped into ‘objective’ measures of awareness (focused on participants’ observable performance on a task) and ‘subjective’ measures (focused on participants’ reports of their own conscious experience). These measures can be used to measure a conscious threshold, though no metric is universally agreed upon for this purpose. We sought to identify an accurate subjective measure and, upon doing so, utilize it in examining key differences between subjective and objective measures. We argue that if a given purported threshold is a threshold of subjective experience, then there is one test that, at a minimum, this threshold must pass: If identical stimuli are presented at two different levels of intensity that straddle the threshold, then subjects should tend to report that the two stimuli are subjectively different. Furthermore, the probability that both stimuli are reported as different should be lower if both stimuli are below or both are above the putative threshold. We estimated thresholds using this approach and compared them to thresholds obtained using objective measures. We examined two things - 1) the alignment of the objective and subjective thresholds in relation to one another and 2) the slope of the fitted psychophysical function for each measure. Preliminary findings provide clear evidence for an abrupt non-linear threshold of conscious perception, and also argue against the existence of ‘blindsight-like’ phenomena in neurologically intact individuals.

The neural correlates of consciousness have long been debated within the scientific community, particularly regarding the involvement of extra-sensory areas, especially the prefrontal cortex (PFC). One key question remains: does it play a role in conscious access itself or merely in task-related activity?

In a recent EEG study (Sergent - Nat Com 2021), Sergent and colleagues conducted single-trial analyses using a simple auditory protocol with near-threshold stimulation under both task-related and task-free conditions. Their findings demonstrated that, in both scenarios, perception is linked to the all-or-none triggering of late activations following initial sensory processing.

In order to explore the underlying networks and their dynamics, we conducted a similar protocol in fMRI. Our results showed that, in both task-related and task-free conditions, stimulus intensity around threshold modulates the activity of a broad network, encompassing both sensory areas and higher-level areas such as the PFC and the anterior insula. However, in the task-free condition, premotor involvement was absent. This aligns with previous EEG source reconstruction and supports our hypothesis of a "global workspace" underlying both conscious access and task-related activity, as well as a "global playground" that includes extra-sensory areas but excludes task-related regions such as premotor areas.

Furthermore, trial-by-trial analysis of activations in these networks allows us to determine which nodes of these networks participate in the all-ore-none dynamics observed in EEG. Finally, multivariate analysis allows a finer understanding of whether and how conscious access with and without a task shapes the encoding of a stimulus in these different regions.

Non-invasive neuroimaging techniques have paved new ways for studying brain connectivity, enabling researchers to map interregional communication. This has fueled a growing interest in subdividing large brain regions based on their distinct connectivity profile to deepen our understanding of the brain’s interconnections. The pulvinar, a thalamic key structure for visual processing, attention, and sensory integration, has become an interesting target for such connectivity-based parcellation studies. Given its extensive connectivity with cortical and subcortical structures, an accurate parcellation of the pulvinar could provide critical insights into its role in higher-order cognitive functions and conscious experience.

Despite significant advancements in human brain connectivity research, there is a need for methods that integrate multiple modalities to refine parcellations. Current approaches typically rely on individual layers of data, limiting their capacity to capture the full complexity of the brain’s organization. To address this, we developed a computational pipeline that combines various layers of connectivity data, including structural (Diffusion Weighted Imaging) and functional (resting-state fMRI) modalities, into a unified annotated connectome. Layers are preliminarily framed as networks and their uniqueness is evaluated exploiting tools from graph theory. A novel application of local o-information is used to combine these into a single “informational connectome”. Clustering algorithms are then applied to this informational layer to identify subdivisions within the pulvinar.

Preliminary results indicate that this method outperforms unimodal parcellation, offering a more accurate representation of the pulvinar's substructural heterogeneity. In this regard, our multimodal approach may contribute to a more comprehensive model of thalamocortical dynamics underlying awareness.

Our experiences do not seem to fit into a scientific world-view, and I have suggestions as to why we find it so difficult to understand consciousness. A two-pronged attack is needed.

First, the difficulty of fitting mental events into our scientific knowledge of the physical can be resolved by accepting my idea that the fundamental object in the universe is an arrangement. These include everything from the strings of basic physics to things not overtly physical.

Second, there are still difficulties when we consider our minds, and I will argue that these arise from our minds' limited communication capabilities.

Our minds have ~10^11 neurons, each firing 0.1 to 200 times per sec, and each with many dendrites and synapses. The specification of the mind for one second in terms of bits of information is difficult to estimate, but must be very much larger than 10^11. The whole activity produces our conscious experience.

There are direct experiences formed from our sense data and indirect experiences triggered by the internal workings of the mind. Our speech has a bit rate that has recently been estimated at only 39 bits/sec, so we cannot describe what is happening in our minds. Nevertheless, speech can activate vivid thoughts and indirect experiences, but these will not include qualia.

We are left in the position of having experiences that we cannot adequately describe. We just have to live with this, and I conclude that the so-called ‘hard problem’ is an ill-posed signpost to a false trail.

Introduction: Consciousness science is at a crossroads, with competing theories, fragmented methodologies, and unresolved foundational questions. In response, we argue for a transdisciplinary approach that transcends traditional academic boundaries and integrates perspectives within and beyond the scientific community. Methods: We introduce a transdisciplinary research design, combining literature review, case studies, mixed-methods interventions, and conceptual analysis to evaluate and extend the limitations of current consciousness studies. We explore collaborative methodologies that bridge disciplines and incorporate perspectives from stakeholders, such as policymakers, practitioners, and community members. We discuss practical examples of ongoing transdisciplinary interventions and offer tools for fostering collaboration that embraces the richness of human experience while advancing the field beyond its current paradigmatic constraints. Results: Our findings suggest that a transdisciplinary approach may foster more comprehensive and socially relevant studies of conscious subjective experience. Case studies include adversarial collaborations, synthetic consciousness, disorders of consciousness, and cross-cultural studies. They illustrate how transdisciplinary collaboration enhances experimental design, ethical considerations, and potential applications. We also propose practical tools for cross-sector collaboration and co-creation of knowledge. Conclusion: Collaboration across sectors—linking neuroscience, social sciences, philosophy, policy and real-world actors—can generate and promote rigorous yet inclusive scientific inquiry and solutions. The approach aligns with the conference themes by emphasising holistic and participatory research strategies to reveal the nature of subjective experience and eventually its correlations with mind-brain-body interactions. This framework also offers a scalable model for scientific integration, shaping future discourse on consciousness in both academic and applied settings.

Debates surrounding the moral significance of consciousness, other than being theoretically important, also have serious practical implications. These debates often heavily rely on assumptions appealing to intuitions about the moral significance of consciousness. However, these assumptions are rarely, if at all, empirically tested. Here, we report the results of an online survey designed to test how prevalent these intuitions actually are. Over a thousand participants answered questions about consciousness, moral status, and their relationship. Our results reveal a complex landscape; while some intuitions were widely shared, others showed considerable variation. For example, we found a strong correlation between attributions of moral status and attributions of consciousness to various entities. However, people’s intuitions diverged with respect to whether consciousness is an all-or-none or a graded phenomenon. Another example of divergent intuitions was found with respect to people’s decisions in moral dilemmas: while respondents’ attitudes can be clustered into meaningful groups, factors other than consciousness seem to explain the difference between the groups. These variations in intuitions depended on features such as demographic background and specific expertise. This work provides valuable insight into how intuitions vary among different populations, shedding light on previously neglected aspects of the debate on the relationship between consciousness and moral status.

In your lifetime, you unceasingly gather evidence suggesting your conscious experiences are congruent with the true state of the world. Sensory inputs of any object usually indicate it is present in your vicinity; when you’re feeling extremely sad, your emotional experience may gain validation through inferential processes; your attitudes, such as opposing drastic reforms in the state’s judicial system, may be conserved and justified by confirmation biases; and your social inferences, such as assuming a feminist bank teller is more likely than just a bank teller, may be reasoned through logical fallacies. Hence, trusting your own conscious experiences to be reliable descriptions of reality seems to be the (mostly implicit) default of our cognitive system. This notion holds significant weight since our behavior—broadly defined to include cognition, motivation, and emotion—can vary drastically based on the perceived veracity of our conscious experiences.

We developed the Trust in Phenomenology Scale (TiPS) and provide cross-cultural evidence that one’s trust in her conscious experiences is consistent across multiple domains (e.g., perceptions, sensations, intuitions, emotions and hunches). We document reliable individual differences, such that some of us consistently trust our conscious experiences more than others. Crucially, our data suggest that trust in conscious experiences has meaningful implications to cognition and behavior, demonstrated by its role in determining perceptual decisions, problem solving and reasoning - utilizing the Cognitive Reflection Test (Frederick, 2005) and the Wason Selection Task (Wason, 1966).

Cognitive phenomenology — chiefly the extent to which occurrent thoughts, such as propositional attitudes, possess distinct phenomenal character — remains a topic of philosophical controversy (e.g. Siewert, 2018; Horgan & Tienson, 2002; Pitt, 2009; Bourget, 2010; Mendelovici, 2018). In my presentation, I will show the significance of this problem for cognitive sciences and argue for an account of Cognitive Phenomenal Unity (CPU).

Even though not all conscious cognition requires a distinct phenomenal format, thought must neccesarily be translatable into an amodal spatial format to interact with perceptual phenomenology, where all modalities possess a spatial component. Such dispositional compatibility is crucial for effective and seamless behavioral coordination and mutual updating between perceptual and cognitive systems. This shared representational format facilitates the projection of cognitive content onto sensory experience, enabling real-time error detection, strategic adjustment, and adaptive learning. To support my claim I will present data from research on apahantasics (Bainbridge et al., 2021; Keogh & Pearson, 2017), conscious error detection (Charles et al., 2013; 2014), and learning (Skóra et al., 2023; 2024).

In the current context, where debates on conscious cognition often downplay phenomenology, this argument underscores the need to re-center phenomenological structures in cognitive models. By aligning cognitive and sensory phenomenology through a unified spatial format, this approach offers a more naturalized account of how conscious cognition informs behavior, and offers new insights into the symbol grounding problem (Harnad, 1990).

We tackle the hard problem of consciousness [1-16] taking the naturally selected, embodied organism as our starting point [17-30]. We provide a formalism [31-41] describing how biological systems self-organise to hierarchically interpret unlabelled sensory information according to valence [42-49]. Such interpretations imply behavioural policies which are differentiated from each other only by the qualitative aspect of information processing. Natural selection favours systems that intervene in the world to achieve homeostatic and reproductive goals. Quality is a property arising in such systems to link cause to affect to motivate interventions [34,50-60]. This produces interoceptive and exteroceptive classifiers and determines priorities [61-69]. In formalising the seminal distinction between access and phenomenal consciousness, we claim that access consciousness at the human level requires the ability to hierarchically model i) the self, ii) the world/others and iii) the self as modelled by others, and that this requires phenomenal consciousness. Phenomenal without access consciousness is likely common, but the reverse is implausible. To put it provocatively: death grounds meaning, and Nature does not like zombies. We then describe the multilayered architecture of self-organisation from rocks to Einstein, illustrating how our argument applies [70-89]. Our proposal lays the foundation of a formal science of consciousness, closer to human fact than zombie fiction.

For references [1-89] see the preprint - Bennett, M. T., Welsh, S., & Ciaunica, A. (2024, June 27). Why Is Anything Conscious?. https://doi.org/10.31219/osf.io/mtgn7

Daniel Dennett's theory of consciousness notably rejects transcendental explanations, viewing the transcendental tradition of self-consciousness as merely a narrative fiction. Instead, Dennett seeks to explain consciousness with methods aligned with natural sciences.

However, critiques of transcendental consciousness theory have existed since the late 18th century(from Jacobi to Henrich), and the theory has evolved significantly in response. This suggests that Dennett's critique does not directly engage with fully developed transcendental theories, such as Fichte’s, but rather with pre-transcendental understanding of consiousness.

Framing transcendental consciousness in terms of Qualia, the very core function of consciousness as synthesis is neglected. Consciousness as a synthetic cognitive faculty is unjustifiably reduced to a mere combination of subjective experiences. To counteract subjective absolutism, Dennett adopts a naturalistic framework, emphasizing functionalist and computational explanations of consciousness. As the culmination of systematic transcendental consciousness theory, Fichte's model not only refutes the critique of narrative fiction but also offers an alternative framework for addressing Chalmers' hard problem of consciousness. According to Fichte, self-consciousness is a reflective process that integrates the external world into the structure of synthetic unity of apperception. Therefore, it is neither the producer of the outside world nor the mystical unthought within thought, but rather the condition that makes cognition—both of ourselves and the world—possible.

The apparent conflict between physicalism and transcendentalism may therefore stem from misunderstanding. While physicalism emphasizes empirical explanations of consciousness, transcendentalism provides a meta-theoretical approach that examines the preconditions for such explanations. In this sense, both can be seen as complementary rather than conflicting perspectives.

Many modern theories of consciousness seek to be consistent with prevailing physical theories. While ontological flexibilities and current uncertainties in physical laws permit a diversity of options, it remains valuable to monitor ideas at the forefront of physical science. Such ideas can generate constraints, implications, or new flexibility for theories seeking to explain the relationship between physical processes and conscious experience.

We present the results of a structured literature review of explicitly ontological physical theories advanced in recent years. We extract 24 distinct theoretical positions from 22 papers shortlisted out of the 182 papers that appear in our 2018–2023 search, providing a recent position on most major schools of thought, especially in quantum mechanics and Quantum Field Theory.

Our findings reveal significant variety across these recent papers. However, we suggest they can be partially captured in their position on eight axes: substrate perspective, property ontology, property plurality, dimensional ontology, dimensional plurality, allowed interactions, world plurality, and intuition rejection (i.e., which intuitions of classical mechanics they reject or maintain). This taxonomy could help consciousness theorists engage more effectively with contemporary debates in physics.

Finally, we outline three specific points of relevance for consciousness theorists: (i) Three strategies for maintaining an “enlightened agnosticism” about physical ontologies; (ii) A default weakening of ontologically-grounded arguments about particular consciousness theories, together with potential methods for re-bolstering those arguments; (iii) An exploitable parallel between physical ontologies and philosophy of mind concerning choices of intuition to preserve, leading to a potential joint endeavour between the disciplines.

Why does the experience of red feel the way it does, rather than like green or the sound of an oboe? The quality space approach, increasingly influential in philosophy and neuroscience, offers a path to answering this question. This approach maps the similarity relations between mental qualities, leading to a model that can be interpreted in two ways: intrinsically (experiences can be similar because they share an intrinsic nature) or relationally (phenomenal character consists in experiencing how one experience is similar to others). The existing literature largely focuses on relationalism, proposing it as a reductionist solution to the problem of phenomenal character. However, relationalism faces serious difficulties, and a thorough discussion of intrinsicalism is missing. This talk aims to articulate the main challenge for relationalism and initiate a discussion on intrinsicalism.

The presentation will begin by introducing quality spaces and the two views, with particular attention to the notion of intrinsicality, which is often misunderstood. I will then examine the main challenge for relationalism: the problem of underdetermination of phenomenal character. Finally, I will propose a model of the substrate of mental qualities, which I call the internally relational model. This model identifies the substrate of mental qualities in specific relations between neural activities that follow a dynamic similar to that of a combustion engine. Although it embraces some of the spirit of relationalism, the model remains intrinsicalist. Moreover, it aligns well with neuroscientific data.

This presentation develops a central argument about qualia—those putative “what-it’s-like” features of experience—and demonstrates that regardless of whether we accept or reject the logical possibility of “zombies,” realism about qualia is severely challenged. Building on recent discussions in analytic philosophy, I introduce a logical “deadlock”:

1. If we accept that zombies (physical duplicates of humans lacking true subjective experience) are possible, we empower a debunking argument, implying our conviction in real qualia could be illusory.

2. If we deny zombies are possible, we must invoke a necessary link between physical and phenomenal properties—but this either forces an untenably reductive stance (whereby qualia are effectively eliminated) or pushes us toward a “mysterious” non-reductive dualism that equally destabilizes classical qualia realism.

By demonstrating that both routes—accepting and denying zombies—undermine a robust realism about phenomenal consciousness, I argue that illusionism becomes the most coherent position. My claim is that the very arguments typically advanced either for anti-physicalism or for refusing zombies’ possibility converge on the conclusion that we do not have “genuine” phenomenal properties. Such a stance has direct bearing on the scientific study of consciousness, urging researchers to probe not for irreducible subjective qualities, but for the cognitive and neural mechanisms underpinning our belief that such qualities exist.

This “deadlock” and its resolution by illusionism speak to the conference’s focus on subjective experience, challenging us to reconsider whether experiential reports track unique non-physical properties, or emerge wholly within the physical dynamics of the mind-brain-body nexus.

Chalmers argues that the existence of suddenly disappearing qualia has two impossible consequences: i) it suggests brute discontinuities in the laws of nature, not found anywhere else; and ii) it implies a specific number of neurons below which qualia would suddenly disappear, which would be arbitrary. Sebastián and Martínez (2024) argue that Chalmers’ argument against the possibility of suddenly disappearing qualia fails, because it falls into a ‘gradualist trap’. They do so by appealing to the phenomenon of bifurcation in dynamical systems. Chalmers’ argument relies on the assumption that stable values of a high-level variable cannot change discontinuously and abruptly in response to gradual changes in a low-level variable. However, a bifurcation is exactly that: an abrupt change in the high-level variable resulting from an infinitesimal change of the low-level one. I discuss the Sebastián and Martínez (2024) argument and conclude that it fails to provide a convincing account against Chalmers’ ultimate conclusion—that is, that fading and suddenly disappearing qualia are both impossible—which is ultimately what we are interested in. In particular, their framework simply demonstrates that the implausibility of suddenly disappearing qualia cannot be based on the claim that such system behaviour is naturalistically unacceptable. At the same time, however, it shows that, if the abrupt loss of qualia is to be understood as a bifurcation-induced transition, it follows that the suddenly disappearing qualia argument can only be resisted so long as there is a possibility for fading qualia.

This paper reframes “moral” responsibility as consequence for action (CFA), a pragmatic tool for behavior modulation, and clarifies the role of consciousness in its application. It argues that while "free" will is incompatible with determinism, its social consequence—responsibility—remains viable under a consequentialist framework.

A decision-making model is introduced, framing humans as input-output systems influenced by internal processes and external inputs. CFA operates as an external regulatory mechanism, designed not for retributive punishment but for maintaining social order.

CFA’s efficacy depends on its ability to influence decision-making. For example, it is ineffective in cases of sleepwalking (automatic behavior) but partially functional in forgetting, where significant consequences prompt proactive measures. Implicit biases further illustrate CFA’s societal context-dependence, as societal norms shape its application.

This account resolves a key critique of consequentialism by linking CFA to decision-making processes. Consciousness, while not defining responsibility, becomes relevant insofar as it affects CFA’s functionality. This reframing challenges traditional moralistic interpretations, emphasizing accountability’s pragmatic role in social living.

Integrated information theory (IIT) is a scientific theory of consciousness that proposes mathematics-based explanations for both the presence vs. absence of experience, as well as for the qualitative nature of experience. Rather than starting from what is known about the behavioral or neural correlates of consciousness (which is more typical among neuroscientific theories of consciousness), IIT takes an axiomatic approach. By starting from introspection, phenomenology, and reason, IIT first identifies a set of essential properties of experience (“axioms”), and then infers corresponding necessary and sufficient conditions for a physical entity to be conscious (“postulates”). There have been many criticisms of IIT, including of its axiomatic framework. In this project, we first review how IIT’s axiomatic framework has evolved from version 1.0 to version 4.0, and why the theory takes an axiomatic approach. We then analyze IIT’s axioms through the lens of the philosophical discipline of phenomenology as it was championed by Edmund Husserl and Martin Heidegger. In our analysis, we describe each of IIT’s axioms, and we provide examples from the writings of Husserl and Heidegger that support or contradict each axiom. We conclude with a discussion of how the field of phenomenology is a rich resource for contemporary theories of consciousness, whether or not they take an axiomatic approach.

The active inference framework proposes that the brain learns an internal model of the world that is used to predict and act upon the distal causes of sensory states. This account describes the internal model as coming into contact with the environment only indirectly, and as such proponents of the theory have acknowledged that conscious perception is secluded from the real environment, ineluctably leading to a form of Cartesian scepticism. Critics have responded with claims that Cartesian scepticism makes the theory self-refuting, or at least unable to fulfil its naturalistic ambitions, making active inference less palatable than alternative theories of conscious perception (such as direct perception accounts). Combining recent literature with insights from C. S. Peirce’s pragmatism, I argue that the apparent problem of Cartesian scepticism occurs due to traditional thinking about representations, which interprets the contents of consciousness as a ‘synchronic’ mapping to the environment. However, the theory is not beholden to this interpretation. Instead, consciousness as the result of a learned internal model is amenable to a diachronic structural conceptualisation of representational vehicle that short-circuits any Cartesian sceptical worries.

Consciousness is often conflated with concepts such as subjectivity, awareness, experience, and sentience. By delineating these distinct yet related terms—starting with consciousness, then exploring information processing and subjectivity—this project aims to clarify the convoluted network of existing terminology. After establishing these definitions, I will examine their presence across various systems, from the foundations of existence to complex organisms, illustrating how awareness, subjectivity, and information processing function as the primary forces of consciousness. Building on the work of Nagel (1974), Block (1995), and Chalmers (2023), as well as insights from molecular and evolutionary biology (Godfrey-Smith, 2019; Levin, 2019) and Eastern traditions such as Advaita Vedanta, this framework refines these concepts for scientific inquiry and communication. Here, consciousness is defined as the integration of awareness, subjectivity, and information processing. Awareness—the essence of consciousness—is the fundamental capacity to detect and respond to phenomena. It is intransitive and unconditional, whereas consciousness is functional, transitive, and relational. Awareness manifests through different systems, shaped by their mode of being (subjectivity) and interaction with the world (information processing). This perspective allows for a 2D classification space, where subjectivity and information processing serve as axes for evaluating potentially conscious systems. With this foundation, we propose precise definitions for related concepts—such as agency, intelligence, cognition, and sentience—disambiguating overlapping terms. Finally, we apply this framework across biological and non-biological entities, from humans and animals to plants, rocks, atoms, and artificial neural networks like ChatGPT.

This research investigates the persistent resistance to illusionism about phenomenal consciousness - the view that our subjective experiences are not real (Frankish, 2016；Kammerer，2021). Through three interconnected analyses, this study examines both the sources and implications of our intuitive rejection of consciousness as an illusion.

First, the research analyses the cognitive and cultural foundations that make denying phenomenal consciousness counterintuitive（Sytsma, J., & Fischer, E.,2023）. By examining how our fundamental assumptions about selfhood and reality shape our resistance to illusionist theories, this work illuminates why we find it difficult to seriously entertain the possibility that our conscious experiences might be illusory.

Second, the study scrutinizes the introspective mechanisms that potentially generate and maintain our conviction in phenomenal consciousness （Daniel Dennett，1991；Derek Pereboom，2009；Elizabeth Irvine 2019；Eric Schwitzgebel，2008）. By critically examining how introspection might systematically misrepresent our mental states, this analysis reveals how our seemingly direct awareness of conscious experience could be constructed rather than discovered through introspection.

The final section extends the illusionist framework beyond human consciousness to evaluate its implications for understanding consciousness in non-human animals and artificial intelligence systems （Umbrello and Sorgner, 2019；Dung，2023）. This analysis develops novel insights into how illusionism might reshape our understanding of consciousness across different types of minds, with significant implications for both theoretical approaches to consciousness and practical ethical considerations.

Scientific research on consciousness and affect is split between two incompatible perspectives. On the one hand, negatively valenced affective states, such as bodily pain, are typically characterised by subjectively felt qualities. On the other hand, some scholars claim that the pre-theoretic, naïve view of pain allows for the existence of unfelt pains. In this ongoing debate, a series of experimental philosophy studies conducted by Reuter and Sytsma (2020) serves as the most significant piece of evidence in support for the latter position. However, it remains unclear whether the results obtained by Reuter and Sytsma are free from biases induced by the design of their questionnaire, which may have inadvertently steered participants toward endorsement of unfelt pains by leaving no room for the possibility of unattended or forgotten pains.

I will present the findings of two online surveys testing the hypothesis that participants do not endorse the existence of unconscious pain when presented with alternative options. The first study is a strict replication of Reuter and Sytsma's studies in a condensed version (N=80), and the second is a modified version of their experimental design (N=94). Our findings clearly show that those of Reuter and Sytsma likely resulted from overlooked ambiguities rather than from the fact that the folk’s intuitive psychology allows for unfelt pains. Based on these results, I defend a view that favours a phenomenal conception of pain, i.e., a view that does not allow pain to exist in absence of someone’s subjective experience of it.

We provide an account of the apparent intermittency of subjective experiences, between conscious and unconscious phases, grounded on three parsimonious notions: (1) the brain is an inference engine and stochastic simulator endowed with a good enough generative model of the world inherited via evolution and forged by experience, (2) the brain’s internal model of the self (itself, its body and actions) is an intermittent and simplified representation invoked only when needed to expedite inference, which specifies and enables first-person subjective experiences through the identification of a model of the self (d-self or synecdoche for the p-self) with the physical self (p-self), and (3) realistic monism as a merger of physicalism and panpsychism. This scheme shifts the focus from problematic standalone subjective experiences to the identification of subjective experiences with the system’s model of itself and its contingent attributes, is consistent with the empirical and phenomenological evidence and provides testable predictions: (1) only macroscopic scale information that is expedient for survival (d-self shell) can become subjective experience, and (2) the hub of subjective experience is mostly distributed along the posterior medial cortex.

I present here the first stage of a larger project on the fundamentality of sensation to consciousness in the wider sense. Some anti-physicalists about consciousness claim that the what-it-is-likeness of experience is inherently mysterious and beyond mechanistic explanation. To evaluate this claim, we need a clear account of the relevant features in need of explanation (mechanistic or otherwise).

I demonstrate how, by taking the phenomenological stance towards ourselves, we can pick out these features. I argue that our sense of our own heart rate, breath rate, location in space, our senses of extensibility and its limits, of being embodied, and our sensation of moving through time, partially describe what it feels like to be me (and you). I aim towards demystifying phenomenal creature consciousness by examining these dynamic, continuous, and yet underestimated sensory processes.

Finally, I suggest that we should rethink the nature of what-it-is-like instead as a pre-cognitive (but fully explainable) phenomenon. First, there are often times when I do not have any explicit cognitive representation of the speed of my heart pumping. There is something it nonetheless feels like to be me at such times which is modified and partially constituted by my having the rate of beating I have. Second, I argue that subjectivity and privileged-access characterizations of phenomenal consciousness are inconsistent with ineffability claims. One way to resolve this disjoint in our intuitions is by allowing that, while a basic component of sensation is a phenomenal feel, that ‘feel’ may sometimes lie beneath our explicit awareness.

How is it that unconscious cognition plays a major role in both heuristic, biased type 1 thinking and in the incubation phase preceding the eureka moment in creative cognition?

To address this question, I combine two traditions within the history of thought concerning the unconscious. One tradition, of which Leibniz is a prominent defender, claims that the mind is cognitively active even when unconscious. Modern dual-process theories that accept this claim use this notion to distinguish between unconscious, automatic type 1, and conscious, analytic type 2 thinking.

The second tradition, as defended by Freud, views consciousness and the unconscious as being in a conflict, emphasising primordial drives in the unconscious. Freud proposes an antithetical semantic for contents of the unconscious, i.e., the meaning of a symbol can denote at the same time one thing and its opposite. Lévi-Strauss adds that the unconscious follows certain structural laws, most notably association.

Using the aforementioned accounts, I explicate the unconscious as that place in the cognitive architecture which transforms elements with antithetical semantics by means of association. This explication accounts for both the role the unconscious plays in creative processes as well as type 1 thinking.

The "inverted qualia" thought experiment posits that qualitatively distinct experiences—such as "red" and "green"—could be systematically swapped across individuals without observable inconsistency. To empirically assess this possibility (see, for example, Palmer, 1999), we propose a mathematical and experimental framework to test the compatibility of two competing mappings of color experience: (A) Non-inverted mapping, where similar colors (e.g., red-to-red) correspond across individuals, and (B) Inverted mapping (any possible swapped mapping not included in (A)), where dissimilar colors (e.g., red-to-green) correspond. To determine the optimal mapping between individuals' color similarity structures, we employ unsupervised alignment using Gromov-Wasserstein Optimal Transport (GWOT) (Kawakita et al., 2024), with the Gromov-Wasserstein Distance (GWD) quantifying structural compatibility—the lower the GWD, the more compatible the mapping. To construct individual color similarity structures, we collected pairwise similarity ratings for 93 colors (4,371 pairs per participant) from 11 participants including both color-neurotypical and atypical individuals. Using full 93×93 similarity matrices, we performed a person-to-person unsupervised alignment. Among color-neurotypicals, we found that the non-inverted mapping predominantly preserved similarity structure, while inverted mappings yielded significantly higher GWD values, indicating lower compatibility. However, for the alignment between color-neurotypical and atypical participants, we observed the opposite: inverted mappings were more compatible than non-inverted mappings. These findings suggest that while subjective experiences are consistent across color-neurotypicals to the extent that inverted mappings are largely incompatible, they may systematically differ between neurotypical and atypical populations, offering an empirical basis for testing the commonalities and variations in subjective experience through an unsupervised structural mapping perspective.

Phenomenal information—the kind of information tied to conscious experience—is inherently subjective, grounded in a point of view. This familiar observation demands precise articulation in informational terms. Unlike objective knowledge, which describes the arrangement of properties in the world, phenomenal information reveals something specific about the subject: the properties they have and their relation to their environment. Known as perspectival or centered-world content, this information reveals one’s location in logical space, offering a uniquely first-person perspective.

A striking feature of conscious experience is its ability to combine sensory data—vision, hearing, and touch—into a unified understanding of the world. At its core, all sensory information serves a shared purpose: providing self-locating information. For example, seeing a table in front of you, hearing a sound to your right, or feeling an object in your hand all contribute to a coherent sense of where you are. Standard models of perception assume each sense processes information in its own frame of reference, requiring computationally demanding translations to integrate them. However, these models struggle to align with empirical data and explain the seamless unity of conscious experience, where sensory information appears perfectly integrated and consistently aligned with the subject’s perspective.

This talk argues that perspectival information provides the key to this unification. By abstracting differences between sensory frames of reference, consciousness integrates sensory data into a coherent, observer-centered perspective. This framework illuminates the connection between self-location and multisensory integration, offering fresh insights into how consciousness organizes the information we use to navigate and understand the world.

Dennett’s intentional stance often leads theorists down one of two paths. The left path leads from Dennett's pragmatism to the widespread attribution of conscious mental states to the simplest organisms (biopsychism). Meanwhile, the right path rejects more-than-pragmatic attributions of such states to any system. It is only ever “as if” something has intentional and phenomenal states (fictionalism). Our two options seem to be cognition all the way down, or metaphors all the way up.

We show this is only the case due to a previous turn - accepting the “common kind assumption” that indistinguishable experiences, whether veridical or illusory, belong to a single kind.

We take the left path when, pace Dennett himself, we construe these common factors as intrinsic properties of conscious experience, ergo qualia. This move turns a harmless (Dennett-friendly?) gradualism about cognition into a form of panpsychism.

We take the right path when the common kind assumption undermines empirical realism. By taking veridical and false experiences to constitute a single kind, the former no longer justifies realist claims. Consequently, mental state attributions are always qualified by the connective “as if”, since to drop it is to make an unjustified, metaphysical claim.

We show that by rejecting the common kind assumption, thereby accepting disjunctivism and perspectival direct realism instead, a path from the intentional stance to the meaningful distinction between truly and only seemingly conscious beings remains open. We argue that Dennett’s own illusionism suggests this was the path he would walk down.

How do objective wavelengths become subjective experience? We propose that color qualia are not fixed sensations but learned associative structures shaped by experience. When we see red, we experience not a raw sensation but an intricate web of associations - all learned through lifetime exposure. This reframing dissolves classic philosophical problems: the inverted spectrum becomes impossible because colors aren't swappable primitives but complex learned structures. More radically, we propose the Mars experiment: a human living on Mars would come to experience their red sky as "blue" - the same quale we associate with Earth's blue sky - because it occupies the same position in their web of daily experiences and associations. In a very real sense, the blue is sky: the color quale is the synthesis of all our experiences with that color. This view is supported by natural statistics experiments showing how our color discriminations emerge from environmental structure. We propose novel perceptual experiments that could expand human color experience beyond trichromatic space through structured exposure to new color associations, demonstrating the potential plasticity of qualia. Modern neural networks provide a mechanism: through hierarchical processing and environmental learning, networks develop internal representations capturing both physical and associative structures of color. This bridges the explanatory gap between stimulus and experience, suggesting consciousness itself might be better understood as structured relationships within learned representational spaces.

I examine two notions of illusionism.

Compatibilist Illusionism (CI): Illusionism is compatible with the existence of phenomenal consciousness (PC).

Incompatibilist Illusionism (II): negation of CI

There are two main arguments:

1. Debunking Argument (DbA)

If beliefs about PC can be explained without PC, then PC is debunked.

Beliefs about PC can be explained without PC.

PC is debunked.

DbA leads to CI.

2. Disqualification Argument (DsA)

If PC exists, then there is direct, infallible knowledge of PC.

There is no direct, infallible knowledge of PC.

PC does not exist.

DsA leads to II.

Premise (2) of DsA relies on an implicit epistemological assumption that knowledge is what appears in judgments or reports. Any such knowledge must be indirect and fallible by definition.

Thus, DsA effectively generalizes too broadly: anything could be said not to exist purely because we lack direct, infallible knowledge of it. But since such knowledge is deemed impossible, that would imply that nothing exists at all. II is incoherent

CI leads to the question “With what exactly is CI compatible?”

PC itself. Self-contradiction.

Supernatural phenomena. But PC is a disanalogy to them.

Infallible, ineffable, private, primitive, directly accessible propperties. It broadens the notion of illusionism.

The failure of II reinforces the intuition that direct knowledge must exist.

Illusionism should be understood as CI. It emphasizes the role of a naturalistic meta-ontological approach.

Two steps to restore PC:

Weakening that approach. The intuition favoring direct knowledge extends beyond this approach.

Demonstrating that direct knowledge requires characteristics of PC.

Does mental entity exist?

This has been a fundamental problem in Western philosophy for hundreds of years (Descartes’ “Cogito, ergo sum” versus Hume’s “Bundle theory”).

However, an original model to reveal the principle of “non-self” in Buddhism which is developed on the basis of both the scientific mechanism behind Buddhist meditation and the empirical materials of Theravada Buddhism solves this problem.

According to Aldous Huxley and Ajahn Brahm, we proposed that the nature of vipassana is enhanced awareness induced in Buddhist meditation, which makes contemplating the five aggregates possible, just like “watching” a slow-motion film.

According to the empirical materials from Theravada Buddhism, if we regard each aggregate as an “awareness” which is the state of being conscious of something, then contemplating the five aggregates would reveal the existence of “awareness of awareness”, and discern that it arises a moment after each aggregate and they do not appear simultaneously.

Thus, it is clear that the notion that there is a constant entity always there knowing or experiencing all aggregates just results from the alternation of “aggregates” and “awareness of awareness”, something that under ordinary conditions happens very quickly. (That's like a torch spinning so fast that it looks like a solid ring of fire exists.)

This would lead to the insight of “non-self”: no subject (or mental entity) of awareness at all.

Obviously, this model bridges Buddhism and Western philosophy, demonstrating that mental entity is just the illusion emerging out of the rapid alternation of “awareness” and “awareness of awareness”.

Intuitions about the ontological status of consciousness play a significant role in investigations of this phenomenon. Some of these intuitions concern phenomenal consciousness. While some researchers hold strong realist intuitions about phenomenal consciousness, others, such as Keith Frankish and François Kammerer, argue that phenomenal consciousness is an illusion. Illusionists claim that the existence of phenomenal consciousness is not self-evident and requires justification. Responses to illusionism, offered by David Chalmers, Martine Nida-Rümelin, and others, often rely on introspective data, asserting that introspection reveals the existence of phenomenal consciousness. Typically, this involves introspection as acquaintance, a specific relation between the self and its conscious states, which is thought to provide direct and infallible access to those states. However, Frankish critiques this approach, arguing that acquaintance is a mystifying relation: not only is the nature of the self unclear but so are the conscious states and the relation between them.

Declan Smithies’ simple theory of introspection offers an alternative, defining introspection as a distinctive way of knowing one’s mental states simply by virtue of being in those states. This theory is elegant, intuitively plausible, and avoids the mysticism Frankish critiques. However, does it provide independent reasons for realism about qualia? I argue that it does not. The simple theory only explains how we know about qualia if they exist; if they do not, it remains applicable to psychological or functional states. Thus, the simple theory of introspection does not independently support realism about qualia, leaving the debate between realism and illusionism unresolved.

Introduction:

The exact appearance and biological foundations of content consciousness during anthropological evolution are unknown. The main hypothesis proposed here is that cumulative cultural evolution, simplified here as tool evolution, was a prerequisite for shaping content consciousness, and that it is thus acquired.

Methods:

The case presented is a theoretical argument based on extensive review of relevant literature from tool and biological evolution, dual inheritance theory, developmental psychology, and neuroscience. The goal was to delineate a comprehensive novel theory enabling experimentally testable hypotheses.

Results:

The concept of the “extended mind” states that external objects are constituents of our cognition by utilizing external operators such as tools. Tool evolution can thus be conceived as an evolution of the “extended mind”. Tool evolution follows Darwinian principles where tools are replicated, modified and selected. Tool evolution has two alternating states: the perceptible material universe of tools in the “tool-sphere” and the instructions to make or use them which are socially transmitted between humans. The accumulation of tools during evolution created intersubjective affordances that synchronize social groups, fostered prosociality, facilitated predictive processing, established a third person perspective, and, ultimately, content consciousness. The entanglement of tool evolution with Darwinian biological evolution led to an adaptation of the human brain to tool evolution.

Conclusion:

Content consciousness cannot be conceived outside of tool evolution and thus has an essential element of social transmission and “uploading”. Any search for a neural correlate of consciousness has to take this into account.

The authors define a partial order among multigraphs to study the relationship

between levels and contents of conscious subjective experience in a unique mathematical setting. Phenomenal structure is understood as extrapolated relations among experiences instead of a set of fixed properties of specific experiences. Our mathematical structure is based on multilayer network theory. The theory of multilayers can be seen as a generalization of network theory, widely used in scientific domains and specifically of most current models of conscious experience (see [1]). From our simple set of assumptions, yet rigorous analysis, we conclude that assuming the comparison and quantification among phenomenal experiences yield only partial comparison, rather than commonly assumed absolute comparability. This has implications for evolutionary and animal consciousness: evolution may encompass diverse modes of experiencing, not necessarily implying larger ones on an absolute scale. Our characterization elucidates structural constraints on experiential comparisons imposed by assumptions and choices made by modellers as active participants in the scientific process. In summary, in light of our phenomenological intuitions, it might be right that some experiences carry qualitative aspects that make them incompatible or non-comparable with other experiences, quantitatively speaking. Some experiences are comparable (e.g. at some experiential levels), but others are not. These results have direct implications for consciousness science, evolution and animal consciousness.

[1] Signorelli, C.M., Boils, J.D. Multilayer networks as embodied consciousness interactions. A formal model approach. Phenom Cogn Sci 23, 1119–1150 (2024). https://doi.org/10.1007/s11097-024-09967-w

Where did we go wrong? After three decades, the field of consciousness studies is still largely mired in the hard problem, assuming that consciousness ‘arises from’ or is ‘generated by’ the brain, and seeking the NCCs. Perhaps the mistake was asking ‘What is it like to be a bat?’ instead of, ‘What is it like to be a bat’s model of the world?’ Unlike Nagel’s question, this question can be answered. It is like whatever a model describes.

If we reject the conventional assumption that physical entities such as bats, people, or machines are conscious and instead say that all the models they construct are conscious, there is no mind/brain split or explanatory gap.

What are these models? Predictive processing theory describes the brain as a hierarchical system in which each level builds models to predict the likely next input from the level below, from simple, fleeting models in early sensory processing to dynamic models of objects, people, and eventually a self. This means that we are models in a controlled hallucination based on predictions in a Bayesian brain.

In shifting how we think about ‘what it is like to be’, this approach provides new directions for research. Instead of searching in vain for the NCCs, neuroscience can investigate how these predictive models change when, for example, anaesthetics block thalamocortical loops, when deep meditation or psychedelics weaken long-range self-related connections, or when their return explains the feeling of ‘waking up’ in lucid dreams.

Creativity is a cornerstone of human evolution and is defined as the multifaceted ability to produce novel and useful artifacts. Although much research has focused on divergent thinking, growing evidence underscores the importance of perceptual processing in fostering creativity, particularly through perceptual flexibility. The present work aims to offer a framework that relates creativity to perception, showing how sensory ambiguity can contribute to the generation of novel ideas. We argue that this perceptual process is fundamental to altered states of consciousness, where shifts in attention, predictive processing, and sensory integration lead to increased sensitivity to perceptual affordances. In doing so, we contextualize the phenomenon of pareidolia, which involves seeing familiar patterns in noisy or ambiguous stimuli, as a key perceptual mechanism of idea generation—one of the central stages of the creative process. We introduce “divergent perception” to describe the process by which individuals engage with the perceptual affordances provided by ambiguous sensory information, and illustrate how this concept could account for the heightened creativity observed in psychedelic and psychotic states. Finally, we discuss future paths for the exploration of divergent perception, including targeted manipulation of stimulus characteristics and the investigation of the balance between bottom-up and top-down cognitive processes. This perspective offers new experimental paradigms for consciousness research by linking phenomenological alterations in perception to underlying neurocognitive processes. By integrating perceptual science with research on psychedelic and psychotic states, this framework provides a foundation for exploring the mechanisms underlying spontaneous creativity, hallucination, and insight across a spectrum of altered states.

How does attention influence gradedness of appearance? Given that perceptual load (an attentional manipulation) influences awareness, we investigated how perceptual load influences gradedness, especially of scene perception. We have already shown that load in a primary task influences the slope of psychometric function obtained from perceptual awareness scale ratings. Here, we aimed to understand the underlying mechanism driven by load that influences scene perception. We manipulated external noise as a means of understanding the potential attentional mechanisms. Participants (N=30) performed scene-categorization and provided visibility-ratings using perceptual awareness scale (PAS) for a background scene (added with any one of seven levels of external noise) while performing letter identification of a target letter among six letters presented in a circular array. In the low-load condition, all the distractors were the same whereas in the high-load condition, distractors were heterogeneous. A four-parameter psychometric model was fitted to both accuracy and awareness ratings. In addition to differences in threshold and slope, there was a significant difference in the lower asymptote (at higher noise levels) for accuracy in the high-load compared to the low-load conditions indicating that high-load influences scene perception via external noise suppression. While there were no differences in asymptotes as a function of load with PAS psychometric curves, there was a significant negative correlation between the differences in the asymptotic parameter of accuracy fits and the differences in the slope of PAS fits between high and load conditions indicating a potential relationship between distractor suppression and gradedness of awareness ratings.

As noted by Zajonc, “We do not just see ‘a house’: we see ‘a handsome house’, ‘an ugly house’, or a ‘pretentious house’”. All perceptions are colored by their valence. According to recent proposals, valence would play an important role in conscious perception (Cleeremans & Tallon-Baudry, 2022) and could influence decision-making (Shenhav, 2024). This suggests the omnipresence of valence in our conscious experience. Thus, even prima facie neutral experiences would have a valence, or a micro-valence (Lebrecht, 2012; Mentec et al. in prep). Here, we aimed at documenting the (dis)similarities between the neural correlates of macro- and micro-valences.

Fifty participants took part in this first EEG study about micro-valences. Pictures of everyday objects or affectively loaded pictures were presented. Participants were asked to passively look at the pictures or to explicitly rate their valence, depending on the block.

Grand mean ERPs highlighted common patterns in the macro and micro conditions in both passive and rating conditions. Despite differences in latency and amplitude, P3 and late positive potential were observed for both micro- and macro-valences, so replicating literature on responses to emotional pictures.

These preliminary results give us first insights into the underlying neural mechanism of micro-valences, which might be closer to macro-valences than put forward in the affective neuroscience literature. Multi-variate pattern analyses will be conducted to further explore the (dis)similarity between macro- and micro-valences. We will look at the decodability of positivity vs negativity in both macro and micro conditions. Results will be presented at the meeting.