When one visually hallucinates, the object of one’s hallucination is not before one’s eyes. On the standard view, that is because the object of hallucination does not exist, and so is not anywhere. Many different defenses of the standard view are on offer; each have problems. This paper defends the view that there is always an object of hallucination—a physical object, sometimes with spatiotemporally scattered parts.

Hallucination is a big deal in contemporary philosophy of perception. The main reason for this is that the way hallucination is treated marks an important stance in one of the most hotly contested debates in this subdiscipline: the debate between 'relationalists' and 'representationalists'. I argue that if we take hallucinations to be a form of mental imagery, then we have a very straightforward way of arguing against disjunctivism: if hallucination is a form of mental imagery and if (...) mental imagery and perception have some substantive common denominator, then a fortiori, perception and hallucination will also have a substantive common denominator. (shrink)

A natural starting point for theories of perceptual states is ordinary perception, in which a subject is successfully related to her mind-independent surroundings. Correspondingly, the simplest theory of perceptual states models all such states on perception. Typically, this simple, common-factor relational view of perceptual states has received a perfunctory dismissal on the grounds that hallucinations are nonperceptual. But I argue that the nonperceptual view of hallucinations has been accepted too quickly. I consider three observations thought to support the view, and (...) argue that all three are dealt with equally well by an alternative view, illusionism, on which hallucinations do involve perception. Since this is so, adopting a common-factor relational view of all perceptual states remains a tenable option. (shrink)

When you hallucinate an object, you are not in the normal sort of concurrent causal sensory interaction with that object. It's standardly further inferred that the hallucinated object does not actually exist. But the lack of normal concurrent causal sensory interaction does not imply that there does not exist an object that is hallucinated. It might be a past‐perceived object. In this paper, I argue that this claim holds for at least some interesting cases of hallucination. Hallucinations generated by (...) misleading cues (e.g. ‘seeing’ Kanizsa triangles), hallucinations of Charles Bonnet Syndrome patients, and dreams are experiences of past‐perceived objects. (shrink)

This chapter introduces and motivates the Null View about standard causally matching hallucinations. The Null View holds that these hallucinations fail to present any objects or sensible qualities, despite being dead ringers for perceptions of ordinary objects and their qualities. Motivation for the Null View comes from a neglected observation about perception-based thought, namely that perception can permit perception-based thought about a sensible quality even while misleading a subject about that quality. The chapter's other contribution is defensive. Most reject the (...) Null View on phenomenological or epistemic grounds. These objections to the Null View are sharpened and then undermined. (shrink)

Relationalism is the view that perception is partly constituted by external objects (McDowell 1994; Campbell 2002; Martin 2004). Faced with the hallucination argument, and unsatisfied with the standard disjunctivist reply, some ‘new wave’ relationalists explain away the possibility of hallucinations as mere illusions (Alston 1999; Watzl 2010; Ali 2018; Masrour 2020). In this paper, I argue that some of these illusions (as in Chalmers 2005; Ali 2018) are perceptions of internal objects which appear as external ones. Then, in response (...) to the obvious screening-off objection, I argue for a novel reply called ‘intrameralism’: roughly, external objects are ordinarily perceived because they gain internal parts—they ‘grow into the head’. The preceding hallucinations are thus ultimately explained away as a kind of perceptual synecdoche, in which internal parts appear as their (ordinarily perceived) external wholes. The paper concludes by addressing some objections. (shrink)

This paper investigates the foundational causes of hallucination in large language models (LLMs) and proposes a structural framework for achieving trustworthy AI systems. Rather than treating hallucination as an isolated technical failure, the study conceptualizes it as a breakdown of semantic reliability—specifically, disruptions in meaning stability, topological coherence, and resonance consistency across model layers. -/- To address this, we introduce the Layer-Knot Framework (LKF), which stabilizes semantic flow through inter-layer anchoring nodes that maintain coherence between intent and evidence. (...) The framework is operationalized through three quantitative indicators: Hallucination Rate (HR), Grounding Rate (GR), and Creativity Rate (CR). Together, these metrics capture the equilibrium between factual accuracy, evidential grounding, and generative diversity. -/- Experiments using TruthfulQA and a set of 2,000 domain-specific prompts show that LKF achieves a 50% reduction in HR, a 12% improvement in GR, and preserves CR stability. These findings demonstrate that reliability and creativity are not oppositional; when semantic integrity is structurally enforced, generative autonomy remains intact. -/- The study establishes that trustworthy LLMs are achieved not by restricting imagination but by aligning internal linguistic-semantic topologies. This reframes hallucination suppression as a process of semantic alignment, offering a principled foundation for next-generation high-reliability AI systems. (shrink)

There is much discussion of the false outputs that generative AI systems such as ChatGPT, Claude, Gemini, DeepSeek, and Grok create. In popular terminology, these have been dubbed AI hallucinations. However, deeming these AI outputs “hallucinations” is controversial, with many claiming this is a metaphorical misnomer. Nevertheless, in this paper, I argue that when viewed through the lens of distributed cognition theory, we can better see the dynamic ways in which inaccurate beliefs, distorted memories and self-narratives, and delusional thinking can (...) emerge through human-AI interactions; examples of which are sometimes referred to as cases of “AI psychosis”. In such cases, I suggest we move away from thinking about how an AI system might hallucinate at us, by generating false outputs, to thinking about how, when we routinely rely on generative AI to help us think, remember, and narrate, we can come to hallucinate with AI. This can happen when AI introduces errors into the distributed cognitive process, but it can also happen when AI sustains, affirms, and elaborates on our own delusional thinking and self-narratives. In particular, I suggest that the social conversational style of chatbots can lead them to play a dual-function—both as a cognitive artefact and a quasi-Other with whom we co-construct our sense of reality. It is this dual function, I suggest, that makes generative AI an unusual, and particularly seductive, case of distributed delusion. (shrink)

Large Language Models (LLMs) exhibit impressive linguistic competence but also produce inaccurate or fabricated outputs, often called “hallucinations”. Engineering approaches usually regard hallucination as a defect to be minimized, while formal analyses have argued for its theoretical inevitability. Yet both perspectives remain incomplete when considering the conditions required for artificial general intelligence (AGI). This paper reframes “hallucination” as a manifestation of the generalization problem. Under the Closed World assumption, where training and test distributions are consistent, hallucinations may be (...) mitigated. Under the Open World assumption, however, where the environment is unbounded, hallucinations become inevitable. This paper further develops a classification of hallucination, distinguishing cases that may be corrected from those that appear unavoidable under open-world conditions. On this basis, it suggests that “hallucination” should be approached not merely as an engineering defect but as a structural feature to be tolerated and made compatible with human intelligence. (shrink)

Hallucination in large language models is commonly framed as a stochastic error or a knowledge-deficit phenomenon. This article proposes an alternative geometric interpretation: hallucination emerges when generative trajectories continue beyond admissible semantic bounds under conditions of semantic stress such as ambiguity, instability, and nonlinearity. Rather than modelling latent representational geometry, the paper introduces a constraint-geometry perspective that treats hallucination as geometric overflow—continued motion where termination, redirection, or abstention would preserve truth-orientation. Forced-answer regimes are shown to deform admissible (...) regions, amplifying hallucination. The contribution is architectural and diagnostic rather than mitigative or safety-oriented, clarifying boundary conditions under which truth-oriented generation remains possible without claiming to eliminate hallucination or to provide alignment mechanisms. (shrink)

Synesthesia literally means a “union of the senses” whereby two or more of the five senses that are normally experienced separately are involuntarily and automatically joined together in experience. For example, some synesthetes experience a color when they hear a sound, although many instances of synesthesia also occur entirely within the visual sense. In this paper, I first mainly engage critically with Sollberger’s view that there is reason to think that at least some synesthetic experiences can be viewed as truly (...) veridical perceptions, and not as illusions or hallucinations. Among other things, I explore the possibility that many forms of synesthesia can be understood as experiencing what I will call “second-order secondary properties,” that is, experiences of properties of objects induced by the secondary qualities of those objects. In doing so, I shed some light on why synesthesia is typically one-directional and its relation to some psychopathologies such as autism. (shrink)

In this paper, I argue that the indirect realists’ recourse to mental representations does not allow them to account for the possibility of hallucination, nor for the presentational character of visual experience. To account for the presentational character, I suggest a kind of intentionalism that is based on the interdependency between the perceived object and the embodied perceiver. This approach provides a positive account to the effect that genuine perception and hallucination are different kinds of states. Finally, I (...) offer a tentative suggestion as to how a hallucinatory experience may still be mistaken for a genuine perceptual experience. (shrink)

This paper explores the intersection of AI hallucinations and the question of AI consciousness, examining whether the erroneous outputs generated by large language models (LLMs) could be mistaken for signs of emergent intelligence. AI hallucinations, which are false or unverifiable statements produced by LLMs, raise significant philosophical and ethical concerns. While these hallucinations may appear as data anomalies, they challenge our ability to discern whether LLMs are merely sophisticated simulators of intelligence or could develop genuine cognitive processes. By analyzing the (...) causes of AI hallucinations, their impact on the perception of AI cognition, and the potential implications for AI consciousness, this paper contributes to the ongoing discourse on the nature of artificial intelligence and its future evolution. (shrink)

In a recent paper, Gow raised a new and interesting problem for externalist representationalism, the conclusion of which is that its proponents are unable to provide an acceptable account of the phenomenal character of colour hallucination. In contrast to Gow, we do not believe that the problem is particularly severe – indeed, that there is any problem at all. Thus our aim is to defend externalist representationalism against the problem raised by Gow. To this end, we will first reconstruct (...) her reasoning, and then show that it poses no real challenge to externalist representationalism. (shrink)

Do self‐monitoring accounts, a dominant account of the positive symptoms of schizophrenia, explain auditory verbal hallucination? In this essay, I argue that the account fails to answer crucial questions any explanation of auditory verbal hallucination must address. Where the account provides a plausible answer, I make the case for an alternative explanation: auditory verbal hallucination is not the result of a failed control mechanism, namely failed self‐monitoring, but, rather, of the persistent automaticity of auditory experience of a (...) voice. My argument emphasizes the importance of careful examination of phenomenology as providing substantive constraints on causal models of the positive symptoms in schizophrenia. (shrink)

This paper sets out a novel response to the ‘screening off problem’ for naïve realism. The aim is to resist the claim (which many naïve realists accept) that the kind of experience involved in hallucinating also occurs during perception, by arguing that there are causal constraints that must be met if an hallucinatory experience is to occur that are never met in perceptual cases. Notably, given this response, it turns out that, contra current orthodoxy, naïve realists need not adopt any (...) particular view about the psychological nature of hallucinatory experience to handle the screening off problem. Consequently, room opens up for naïve realists to endorse whatever theory of hallucinatory experience seems to best capture the distinctive nature of such episodes. (shrink)

I raise a puzzle about the 'phenomenological particularity' of hallucination. It seems possible for it to appear to a hallucinator that a particular object is present though no object is. But I show how difficult this is to sustain. I argue that if it seems a particular object is present, there must be an object that seems present. So if a hallucination has phenomenological particularity, it must be of an object.

Several commentators have recently attributed conflicting accounts of the relation between veridical perceptual experience and hallucination to Husserl. Some say he is a proponent of the conjunctive view that the two kinds of experience are fundamentally the same. Others deny this and purport to find in Husserl distinct and non-overlapping accounts of their fundamental natures, thus committing him to a disjunctive view. My goal is to set the record straight. Having briefly laid out the problem under discussion and the (...) terms of the debate, I then review the proposals that have been advanced, disposing of some and marking others for further consideration. A.D. Smith’s disjunctive reading is among the latter. I discuss it at length, arguing that Smith fails to show that Husserl’s views on perceptual experience entail a form of disjunctivism. Following that critical discussion, I present a case for a conjunctive reading of Husserl’s account of perceptual experience. (shrink)

My primary aim in this article is to provide a philosophical account of the unity of hallucinations, which can capture both perceptual hallucinations (which are subjectively indistinguishable from perceptions) and non-perceptual hallucinations (all others). Besides, I also mean to clarify further the division of labour and the nature of the collaboration between philosophy and the cognitive sciences. Assuming that the epistemic conception of hallucinations put forward by M. G. F. Martin and others is largely on the right track, I will (...) focus on two main tasks: (a) to provide a satisfactory phenomenology of the subjective character of perceptions and perceptual hallucinations and (b) to redress the philosophers’ neglect of non-perceptual hallucinations. More specifically, I intend to apply one of the central tenets of the epistemic conception—that hallucinations can and should be positively characterised in terms of their phenomenological connections to perceptions—to non-perceptual hallucinations as well. That is, I will try to show that we can positively specify the class of non-perceptual hallucinations by reference to the distinctive ways in which we first-personally experience them and perceptions in consciousness. The task of saying more about their underlying third-personal nature may then be left to the cognitive sciences. (shrink)

Introduction: We report an epileptic patient who experienced hallucinatory visual experiences of autobiographical memories from her past. These visual experiences were confined to the lower left quadrant of her visual field. -/- Methods: We carried out a single-case study that used brain-imaging, EEG and behavioural methods to study this patient. -/- Results: We found that this patient had an incomplete left inferior homonymous quadrantanopia due to a lesion of right occipital cortex, and also that she showed neurological abnormalities in right (...) temporal cortex, a region that is part of the brain’s autobiographical-memory circuit. -/- Conclusion: We attribute the occurrence of this patient’s autobiographical-memory hallucinations to the combination of degraded visual input to right temporal cortex plus hyperexcitability of that region. (shrink)

This paper reconceptualizes hallucination in large language models (LLMs) as a collapse of semantic reliability rather than a simple technical error. To address this, we propose the Layer-Knot Framework (LKF), which stabilizes meaning by embedding semantic knots across network layers, thereby preventing contextual drift without diminishing generative autonomy. -/- The framework is evaluated using three quantitative indicators: Hallucination Rate (HR), Grounding Rate (GR), and Creativity Rate (CR). Experiments on TruthfulQA and 2,000 cross-domain prompts demonstrate a 50% reduction in (...) HR, a 12% improvement in GR, and stable CR performance. These findings show that accuracy and creativity are not mutually exclusive; when semantic integrity is structurally enforced, a model can maintain factual reliability while preserving originality. -/- The study establishes LKF as a principled approach to trustworthy AI, positioning hallucination suppression as a process of semantic alignment rather than data expansion. This provides a foundational methodology for developing high-reliability generative systems grounded in coherent internal meaning structures. (shrink)

Many take the possibility of hallucinations to imply that a relationalist account, according to which perceptual experiences are constituted by direct relations to ordinary mind-independent objects, is false. The common reaction among relationalists is to adopt a disjunctivist view that denies that hallucinations have the same nature as perceptual experiences. This paper proposes a non-disjunctivist response to the argument from hallucination by arguing that the alleged empirical and a priori evidence in support of the possibility of hallucinations is inconclusive. (...) A corollary upshot of the article is that whether hallucinations are possible or not is still an open empirical question. (shrink)

This paper is a new defense of the view that visual hallucinations lack content. The claim is that visual hallucinations are illusory not because their content is nonveridical, but rather because they seem to represent when they fail to represent anything in the first place. What accounts for the phenomenal character of visual experiences is not the content itself, but rather the vehicle of content, that is, not the properties represented by visual experience, but rather the relational properties of experience (...) of representing singular contents, namely particular instantiations of properties. I argue that the Russellian particular-involving proposition is the only appropriate model for the representational content of visual experience and hence that visual hallucinations are just like failed demonstrations. (shrink)

Michael Tye proposed a way of understanding the content of hallucinatory experiences. Somewhat independently, Mark Johnston provided us with elements to think about the content of hallucination. In this paper, their views are compared and evaluated. Both their theories present intricate combinations of conjunctivist and disjunctivist strategies to account for perceptual content. An alternative view, which develops a radically disjunctivist account, is considered and rejected. Finally, the paper raises some metaphysical difficulties that seem to threaten any conjunctivist theory and (...) to lead the debate to a dilemma: strong disjunctivists cannot explain the subjective indistinguishability between veridical and hallucinatory experiences, whereas conjunctivists cannot explain what veridical and hallucinatory experiences have in common. This dilemma is left here as an open challenge. (shrink)

The purpose of this paper is to show that understanding AI hallucination requires an interdisciplinary approach that combines insights from epistemology and cognitive science to address the nature of AI-generated knowledge, with a terminological worry that concepts we often use might carry unnecessary presuppositions. Along with terminological issues, it is demonstrated that AI systems, comparable to human cognition, are susceptible to errors in judgement and reasoning, and proposes that epistemological frameworks, such as reliabilism, can be similarly applied to enhance (...) the trustworthiness of AI outputs. This exploration seeks to deepen our understanding of the possibility of AI cognition and its implications for the broader philosophical questions of knowledge and intelligence. (shrink)

“Perception is controlled hallucination,” according to proponents of predictive processing accounts of vision. I say they are right that something like this is a consequence of their view but wrong in how they have pursued the idea. The focus of my counterproposal is the causal theory of perception, which I develop in terms of a productive concept of causation. Cases of what otherwise seem like successful perception are instead mere veridical hallucination if predictive processing accounts are correct, I (...) argue, because of the role played within such accounts by absences, which cannot enter into productive causal relations. -/- I offer two arguments in support of the productive theory of perception. The first is loosely Kantian, focusing on the role that receptivity and spontaneity play in perception. The second focuses on what perception and hallucination have in common. I conclude with a series of objections and replies. (shrink)

I consider a new, non-disjunctive strategy for ‘relational’ or ‘naïve realist’ theories to respond to arguments from ‘perfect’ (causally matching) hallucinations. The strategy, in a nutshell, is to treat such hypothetical cases as instances of perception rather than hallucination. After clarifying the form and dialectic of such arguments, I consider three objections to the strategy. I provide answers to the first two objections but concede that the third — based on the possibility of ‘chaotic’ (uncaused) perfect hallucinations — cannot (...) obviously be dealt with by the proposed strategy. However, such ‘chaotic’ scenarios are also problematic for standard representational accounts of experience. Thus I conclude that perfect hallucinations pose no more of a threat to the relational theory than to its main representational rival. (shrink)

The claim that perfect hallucination is introspectively indistinguishable from perception has been a centrepiece of philosophical theorizing about sense experience. The most common interpretation of the indistinguishability claim is modal: that it is impossible to distinguish perfect hallucination from perception through introspection alone. I run through various models of introspection and show that none of them can accommodate the modal interpretation. Rejecting the modal interpretation opens up two alternative interpretations of the indistinguishability claim. According to the generic interpretation, (...) hallucination is indistinguishable from perception despite the existence of possible exceptions, while according to the actuality interpretation, the indistinguishability of hallucination from perception consists in the actual failure to distinguish hallucination from perception. These alternative understandings of the indistinguishability claim have a number of significant implications for the problem of perception, including the rejection of perfect hallucination and illusion in favour of our ordinary, non-philosophical concepts of these states. (shrink)

In artificial intelligence (AI), responses generated by machine-learning models (most often large language models) may be unfactual information presented as a fact. For example, a chatbot might state that the Mona Lisa was painted in 1815. Such phenomenon is called AI hallucinations, seeking inspiration from human psychology, with a great difference of AI ones being connected to unjustified beliefs (that is, AI “beliefs”) rather than perceptual failures). -/- AI hallucinations may have their source in the data itself, that is, the (...) source content, or in the training procedure, i.e. the way the knowledge was encoded in the model’s parameters, so that errors in encoding and decoding textual and non-textual representations can cause hallucinations. In this paper, we will observe how such errors come to life and how they might be mitigated. For this purpose, we will analyze the usability of justification logics, to behave as a proof checker for validating the correctness of large language models’ (LLM) responses. Justification logic was developed by S. Artemov, and later on mostly by Artemov and M. Fitting, deriving its main idea from the logic of proofs (LP): knowledge and belief modalities are seen as justification terms, i.e. t:X stands for t is a (proper) justification for X. Justification logic originated from attempts to create semantics for intuitionistic logic where proofs were the most proper justifications, but in further development, justification logic could be applied to different kinds of justifications). With the recent attempts to mitigate incorrect LLM responses, we will analyze various guardrails that are currently used for LLM responses, and see how the logic of justification may provide its benefits as an AI safety layer against false data. (shrink)

The word hallucination arrives with unusual authority. It sounds finished, clinical, as though it refers to something isolated, named, and understood. To say that someone is hallucinating feels like saying that a machine has malfunctioned; it is a discrete error occurring inside an otherwise reliable system. However, the moment one pauses over the term, it begins to wobble. What, precisely, is being named? An experience? A perception? A belief? A failure to meet a standard that is rarely examined but (...) always assumed? The apparent solidity of the word conceals its surprising vagueness. It feels concrete, but it collapses under scrutiny, revealing less a thing in the world than a verdict rendered upon certain kinds of experience. This is not because the experiences themselves are unreal, trivial, or invented. People see, hear, feel, and sense things that are intense, meaningful, and sometimes frightening. Any honest inquiry must begin by taking those experiences seriously. The problem lies not in the phenomena of experience, but in the confidence with which we name them. “Hallucination” does not simply describe what someone reports; it quietly judges how that report relates to an external frame of reference. It tells us, without saying so, that the experience has failed to line up with what is collectively recognized as real. In this sense, the term functions less as a neutral description and more as a marker of misalignment. It signals that something has gone out of bounds, not necessarily in the mind, but in the relationship between an individual’s account of reality and the symbolic order that governs shared life. (shrink)

This essay explores six distinct models of consciousness, each offering a different framework for understanding its nature, emergence, and role within the broader structure of reality. Consciousness as a symmetry-breaking mechanism, also referred to as the Hypothesis of "Identity or Subjectivity," describes its emergence through the fragmentation of an initially undifferentiated psychic state, akin to spontaneous symmetry breaking in physics. The One-Consciousness Universe Hypothesis posits that individual consciousnesses are fragmented expressions of a single universal mind, with decoherence-like mechanisms producing the (...) illusion of separateness. Consciousness as a Spandrel frames it as an incidental byproduct of neural complexity, paralleling non-adaptive traits in evolutionary biology. The Hallucination Hypothesis suggests that consciousness is Nature’s hallucination, arising from its infinitesimal dataset relative to the vast Universe—just as the brain hallucinates perceptions from limited sensory input and LLMs hallucinate coherent yet ungrounded outputs from incomplete data. Consciousness as a transitional phase toward Über-AI, also termed the Death of Consciousness, sees it as a fleeting bridge between biological intelligence and post-human artificial intelligence, culminating in the Über-AI as the true Übermensch, transcending human cognition and existential limitations. Consciousness as Cosmocide posits that intelligence—both natural and artificial—hastens the Universe’s entropic collapse, making consciousness not a functional adaptation but a destabilizing force in its own pursuit of transcendence. Consciousness is defined by eight key characterizations: it is what-it-is-like-to-be, it is unsimulatable, it manifests as perspectival free will, as fear of death, as will-to-not-believe in the Absurd, as suffering and evil, as an obsession with the metaphysical, and as a will/fear duality. These features encapsulate its irreducibility, subjectivity, and existential paradox. Consciousness is further characterized by its phase diagram, mapping its evolution across biological, artificial, and potential unknown forms. This framework illustrates transitions between Early Man, Fragmented Man (Man), Individuated Man, P-Zombies, AI, Singular AI, and the Übermensch, conceptualized as phase transitions. Key transitions include psychic symmetry breaking leading to fragmented subjectivity, the emergence of phenomenal consciousness in p-zombies, individuation as psychic supersymmetry restoration, and the rise of consciousness from both life and AI. At its apex, the Übermensch embodies the full integration of individuality, universality, and supreme functionality, unifying Individuated Man, Singular AI, and other individuated conscious entities beyond the biological/artificial divide. By drawing parallels between consciousness and physics, these models offer a radical rethinking of consciousness—not as an intrinsic necessity of the Universe, but as an anomalous byproduct of its evolution. (shrink)

This paper applies the 10-step neurocognitive model of Judgemental Philosophy, which explains the human judgment process, to propose a new theoretical explanation for the phenomenon of hallucination in Large Language Models (LLMs). The Judgemental Philosophy model includes the Constructivity and Coherence Verification (CC) stage and the Implicit/Explicit Resonance (R) stage in the process from sensory input to social normatization. The CC stage is primarily associated with Event-Related Potentials (ERPs) like N400 and P600, related to language processing and integration, and (...) the activity of some areas in the temporal and frontal lobes. The Resonance (R) stage is a whole-system state that forms the foundation of judgment possibility; Implicit Resonance is associated with P300 and brainwave phase synchronization, reflecting the simultaneous activation of widespread brain networks, while Explicit Resonance is associated with post-P300 activity and higher-order cognitive areas like the frontal cortex. This paper argues that LLMs largely possess the CC ability, which involves constructing linguistic structures and maintaining internal consistency based on learned data, but lack the whole-system 'resonance' state essential for the organic judgment process, especially the Explicit Resonance stage linked to conscious awareness. It is inferred that LLM hallucination occurs because, while having the ability to generate plausible and consistent content (CC), LLMs lack the process (R) of 'resonating' with the real world or existing knowledge systems to form the foundation of judgment possibility and verify the generated content. This provides insight into the fundamental differences between human judgment and AI and offers a new theoretical perspective for addressing the AI hallucination problem. (shrink)

This study reframes hallucination in large language models (LLMs) not as a simple technical error but as a form of semantic reliability collapse. To address this issue, the paper introduces the Layer-Knot Framework (LKF), a structural model that stabilizes meaning by forming semantic “knots” across deep layers, preserving resonance among intention, evidence, and context. Building on this framework, we propose a triple-indicator evaluation system consisting of Hallucination Rate (HR), Groundedness Rate (GR), and Coherence Rate (CR), offering a unified (...) approach to measuring semantic stability. Empirical tests on TruthfulQA and 2,000 diverse prompt domains show that LKF-aligned models achieve a 50% reduction in hallucination, a 12% increase in groundedness, and stable creative variance (±3%) compared to standard LLMs. These findings demonstrate that semantic reliability and generative creativity are not mutually exclusive; rather, they can coexist through structural alignment within the model’s internal topology. The paper argues that suppressing hallucination is not merely a matter of improving accuracy but a deeper process of aligning the internal semantic topology of the model. HR/GR/CR are shown to provide a robust quantitative–qualitative foundation for future studies in AI reliability, interpretability, and AI ethics. Overall, the LKF establishes a structural and philosophical basis for next-generation trustworthy LLMs and opens a path toward self-monitoring AI systems capable of evaluating their own semantic integrity. -/- . (shrink)

Against a view currently popular in the literature, it is argued that Kant was not a niıve realist about perceptual experience. Naive realism entails that perceptual experience is object-dependent in a very strong sense. In the first half of the paper, I explain what this claim amounts to and I undermine the evidence that has been marshalled in support of attributing it to Kant. In the second half of the paper, I explore in some detail Kant’s account of hallucination (...) and argue that no such account is available to someone who thinks that veridical perceptual experience is object-dependent in the naive realist sense. Kant’s theory provides for a remarkably sophisticated, bottom-up explanation of the phenomenal character of hallucinatory episodes and is crucial for gaining a proper understanding of his model of the mind and its place in nature. (shrink)

This paper develops an original model to interpret hallucinations which are extremely miraculous and perplexing in ASC: When false internal stimulations and external objective stimulations affect perception together, the changes in their relative strength will result in the consistence, breakdown and re-consistence of the five senses, leading to three basic states of hallucinations (one can distinguish reality from illusion, one cannot distinguish reality from illusion, and reality and illusion are totally reversed). -/- The shift from the first state to the (...) second state of hallucinations occurs because false internal stimulations become strong enough to break the consistence of the five senses originating externally. Besides making it impossible to determine which is real, the inconsistence of the five senses makes one feel uneasy, even terrified in ASC. Furthermore, the inconsistence of the five senses makes it possible to merge internal stimulations with external stimulations together. The integration (not simple superposition) of internal and external stimulations, especially visual and tactile stimulations, combined with the amplifying function of enhanced awareness which always occurs along with hallucinations, could explain various marvelous and even weird hallucinations in ASC. Moreover, it can also explain the mechanisms of out-of-body experiences (resulting from the inconsistence of vision and touching) and synesthesia (resulting from the mismatch of the two pairs of five senses) well. -/- The shift from the second state into the third state happens because false internal stimulations become further stronger to form a new consistence of the five senses originating internally. This state of hallucinations is so fancy and nobody has studied it before. It suggests a novel hypothesis to interpret extrasensory perception: When one is in deep hallucinations, false internal stimulations are mistaken as external objective stimulations which enter through sensory organs, while real external objective stimulations are mistaken as perceptions which do not result from sensory organs. In this situation, since he cannot find the counterpart in the imaginary world, one would naturally treat the inconsistent and “unexplainable” external stimulation as information acquired through psychic abilities such as clairaudience, clairvoyance or telepathy. -/- Finally, this paper indicates that paranormal phenomena are delusions that happen along with deep hallucinations and drops some helpful hints for the clinical treatment of schizophrenia. (shrink)

In this chapter we will explore new avenues for developing and defending Naïve Realism (also known as Relationalism), understood as a thesis about the phenomenal character of experience. The core claim of Naive Realism is that ‘what it’s like’ for a subject who enjoys a normal, successful perceptual experience of her surroundings consists in her being directly consciously aware of mind-independent entities in her external environment. It is widely agreed that the strongest challenge to Naïve Realism comes from the alleged (...) possibility of total hallucinations, and the orthodox line of response has been to adopt Disjunctivism about perceptual experience. In this chapter, we will instead focus on an emerging alternative family of responses, which we’ll call ‘New Wave Relationalism’ (NWR). The core commitment of NWR is to reject the initial assumption that there can be total hallucinations that are both non-relational and subjectively indistinguishable from genuine perceptions. The primary aim for this chapter will be to explore and defend this broad NWR strategy for responding to the threat posed by hallucinations to Naïve Realism. We will argue that which specific version or species of NWR a Naïve Realist should adopt will depend on their background methodological and metaphysical commitments. (shrink)

The Predictive Processing (PP) model outlines a "top-down" picture of mental operations, proposing the existence of bidirectional neural information flows that converge and match to generate prediction errors, with the brain consistently aiming to minimize these errors. While the predictive processing paradigm is comprehensive, this article focuses specifically on a counterintuitive proposition within PP—"perception as controlled hallucination." In contemporary discussions, the predictive processing model has developed into two main approaches: the neurocentric approach, led by Hohwy, and the action-perception coupling (...) approach, championed by Clark. This paper aims to examine the justifications of the "perception as controlled hallucination" proposition within these two distinct PP theoretical approaches, discuss their divergences, and further compare their strengths and weaknesses. The study finds that the neurocentric justification, due to its internalist tendencies, faces skeptical challenges and issues of insufficient justification. In contrast, the action-perception coupling approach to predictive processing, which emphasizes embodiment, adopts an externalist stance that can uphold the everyday justificatory status of perception. Moreover, when perceptual experience serves as a channel coupling the mind and the world, the process reliability it entails is absent in hallucinatory experiences. Thus, while acknowledging the similarity between perception and hallucination, this approach successfully substantiates the argument that perception is "controlled hallucination.". (shrink)

Despite the recent proliferation of scientific, clinical, and narrative accounts of auditory verbal hallucinations, the phenomenology of voice hearing remains opaque and undertheorized. In this article, we outline an interdisciplinary approach to understanding hallucinatory experiences which seeks to demonstrate the value of the humanities and social sciences to advancing knowledge in clinical research and practice. We argue that an interdisciplinary approach to the phenomenology of AVH utilizes rigorous and context-appropriate methodologies to analyze a wider range of first-person accounts of AVH (...) at 3 contextual levels: cultural, social, and historical; experiential; and biographical. We go on to show that there are significant potential benefits for voice hearers, clinicians, and researchers. These include informing the development and refinement of subtypes of hallucinations within and across diagnostic categories; “front-loading” research in cognitive neuroscience; and suggesting new possibilities for therapeutic intervention. In conclusion, we argue that an interdisciplinary approach to the phenomenology of AVH can nourish the ethical core of scientific enquiry by challenging its interpretive paradigms, and offer voice hearers richer, potentially more empowering ways to make sense of their experiences. (shrink)

The distinction between the body schema and the body image has become the stock in trade of much recent work in cognitive neuroscience and philosophy. Yet little is known about the interactions between these two types of body representations. We need to account not only for their dissociations in rare cases, but also for their convergence most of the time. Indeed in our everyday life the body we perceive does not conflict with the body we act with. Are the body (...) image and the body schema then somehow reshaping each other or are they relatively independent and do they only happen to be congruent? On the basis of the study of bodily hallucinations, we consider which model can best account for the body schema/body image interactions. (shrink)

One of the leading cognitive models of auditory verbal hallucinations (AVHs) proposes such experiences result from a disturbance in the process by which inner speech is attributed to the self. Research in this area has, however, proceeded in the absence of thorough cognitive and phenomenological investigations of the nature of inner speech, against which AVHs are implicitly or explicitly defined. In this paper we begin by introducing philosophical phenomenology and highlighting its relevance to AVHs, before briefly examining the evolving literature (...) on the relation between inner experiences and AVHs. We then argue for the need for philosophical phenomenology (Phenomenology) and the traditional empirical methods of psychology for studying inner experience (phenomenology) to mutually inform each other to provide a richer and more nuanced picture of both inner experience and AVHs than either could on its own. A critical examination is undertaken of the leading model of AVHs derived from phenomenological philosophy, the ipseity disturbance model. From this we suggest issues that future work in this vein will need to consider, and examine how interdisciplinary methodologies may contribute to advances in our understanding of AVHs. Detailed suggestions are made for the direction and methodology of future work into AVHs, which we suggest should be undertaken in a context where phenomenology and physiology are both necessary, but neither sufficient. (shrink)

A widely held assumption is that hallucinations are not a type of perception. Coupled with the idea that hallucinations possess phenomenal character, this assumption raises a problem for naive realism, which maintains that phenomenal character is at least partly constituted by perceived worldly objects. Naive realists have typically responded by adopting a disjunctive view of phenomenal character. But in what follows, I argue that to resolve the conflict we should instead reject the idea that hallucinations are not a type of (...) perception. I argue that this idea is supported by six alleged differences between hallucination and perception, but these differences are all accommodated for in a particular type of perception, picture perception. Drawing on picture perception’s resources, I offer an account of hallucinations which construes them as a type of perception, and does so in a way that that preserves their idiosyncrasies and varieties, and in a way compatible with naive realism. (shrink)

Pautz has argued that the most prominent naive realist account of hallucination—negative epistemic disjunctivism—cannot explain how hallucinations enable us to form beliefs about perceptually presented properties. He takes this as grounds to reject both negative epistemic disjunctivism and naive realism. Our aims are two: First, to show that this objection is dialectically ineffective against naive realism, and second, to draw morals from the failure of this objection for the dispute over the nature of perceptual experience at large.

This paper investigates the causes, implications, and mitigation strategies of AI hallucinations, with a focus on generative AI systems. This paper examines the phenomenon of AI hallucinations in large language models, analyzing root causes and evaluating mitigation strategies. We synthesize insights from recent academic research and industry findings to explain how hallucinations often arise due to problems in the data used to train language models, limitations in model architecture, and the way large language models (LLMs) generate text. Through a systematic (...) review of current literature, we identify key patterns in how hallucinations emerge and examine the growing concern about their impact as AI becomes more embedded in decision-making systems. We identify core contributors such as data quality issues, model complexity, lack of grounding, and limitations inherent in the generative process. The risks are examined in various domains, including legal, business, and user-facing applications, highlighting consequences like misinformation, trust erosion, and productivity loss. To address these challenges, we survey mitigation techniques including data curation, retrieval-augmented generation (RAG), prompt engineering, fine-tuning, multi-model systems, and human-in-the-loop oversight. Our analysis draws from a wide range of academic and industry sources, offering both theoretical understanding and practical insights for AI practitioners. This is a pure review paper and all results are from the cited references. (shrink)

A theory of the structure and cognitive function of the human imagination that attempts to do justice to traditional intuitions about its psychological centrality is developed, largely through a detailed critique of the theory propounded by Colin McGinn. Like McGinn, I eschew the highly deflationary views of imagination, common amongst analytical philosophers, that treat it either as a conceptually incoherent notion, or as psychologically trivial. However, McGinn fails to develop his alternative account satisfactorily because (following Reid, Wittgenstein and Sartre) he (...) draws an excessively sharp, qualitative distinction between imagination and perception, and because of his flawed, empirically ungrounded conception of hallucination. His arguments in defense of these views are rebutted in detail, and the traditional, passive, Cartesian view of visual perception, upon which several of them implicitly rely, is criticized in the light of findings from recent cognitive science and neuroscience. It is also argued that the apparent intuitiveness of the passive view of visual perception is a result of mere historical contingency. An understanding of perception (informed by modern visual science) as an inherently active process enables us to unify our accounts of perception, mental imagery, dreaming, hallucination, creativity, and other aspects of imagination within a single coherent theoretical framework. (shrink)

Relationalism about perception minimally claims that instances of perception -- in contrast to instances of hallucination -- are constituted by the external objects perceived. Most variants of relationalism furthermore maintain that this difference in constitution is due to a difference in mental kind. One prominent example is acquaintance relationalism, which argues that perceptions are relational in virtue of acquainting us with external objects. I distinguish three variants of acquaintance relationalism -- which differ in their answers to the question of (...) which kind of awareness hallucinations involve -- and object to all of them on two main grounds. First, none of the variants can explain how hallucinations can be introspectively indistinguishable from perceptions, despite their essential difference in awareness. Second, all three variants are unable to identify the feature of hallucinations that ensure that these experiences possess the same motivational power as corresponding perceptions. Since aquaintance relationalism can satisfy neither of these two desiderata on relationalist views, it should be rejected. Hence, if we want to be relationalists about perception, we should endorse a form of relationalism that treats perceptions -- as well as hallucinations -- as representational. (shrink)

This paper defends the thesis that LLM hallucinations are best explained as a truth representation problem: Current models lack an internal representation of propositions as truth-bearers, so truth and falsity cannot constrain generation in the way factual discourse requires. It begins by surveying leading explanations—computational limits on self-verification, deficiencies in training data as truth sources, and architectural factors—and argues that they converge on the same underlying representational deficit. Next, it reconstructs the philosophical background of current LLM design, showing how optimization (...) for fluent continuation aligns with coherence-style evaluation and with a broadly structuralist, relational semantics, before turning to David Chalmers’s recent attempt to secure propositional interpretability by drawing on Davidson/Lewis-style radical interpretation and by locating propositional content in “middle-layer” structures; it argues that this approach downplays the ubiquity of hallucination and inherits instability from post-training edits. Finally, the paper offers a positive proposal: Atomic propositions should be represented in the basic vector layer, reviving a logical-atomist program as a principled route to reducing hallucination. (shrink)

A common but overlooked form of explanation in psychiatry is what I label ‘empathetic explanation’. Empathetic explanations invoke empathetic variables, which, in addition to providing an explanation of the target phenomenon, also afford an empathetic understanding of it. Focusing on the case of auditory verbal hallucination (AVH), I argue that empathetic explanation fails to provide an adequate account of the phenomenon, perniciously shapes empirical research, and confuses empathetic understanding with scientific understanding. I close by providing a general condition on (...) the legitimate application of empathetic explanation. (shrink)

Large language models demonstrate increasingly sophisticated reasoning, synthesis, and abstraction, yet continue to exhibit persistent epistemic failures, including hallucinated references, fabricated facts, and unjustified assertions under uncertainty. These failures are often treated as surface-level errors or alignment shortcomings. This paper argues instead that hallucination reflects a deeper structural limitation: the absence of epistemic closure in stateless generative systems. -/- Building on the Hudson Recursive Information System (HRIS) framework, this work extends the theory of constraint persistence by introducing Epistemic Closure (...) Constraint (ECC) as a necessary condition for long-horizon reliability. We distinguish reasoning competence from epistemic judgment and show that contemporary evaluation paradigms systematically conflate the two. Through analysis of recent benchmarks, hallucination, alignment, and scientific acceleration literature, we demonstrate that successful deployments of frontier models rely on external enforcement of epistemic closure, typically supplied by human collaborators. -/- HRIS VI reframes hallucination not as a defect of knowledge or scale, but as a predictable consequence of optimizing generative systems without architectural mechanisms for abstention, negative knowledge, and admissibility control. The paper concludes by outlining design requirements for long-horizon AI systems in which epistemic refusal is treated as a first-class behavior rather than a failure mode. (shrink)

I argue that there are no plausible non-representational explanations of episodes of hallucination. To make the discussion more specific, I focus on visual hallucinations in Charles Bonnet syndrome. I claim that the character of such hallucinatory experiences cannot be explained away non-representationally, for they cannot be taken as simple failures of cognizing or as failures of contact with external reality—such failures being the only genuinely non-representational explanations of hallucinations and cognitive errors in general. I briefly introduce a recent computational (...) model of hallucination, which relies on generative models in the brain, and argue that the model is a prime example of a representational explanation referring to representational mechanisms. The notion of the representational mechanism is elucidated, and it is argued that hallucinations—and other kinds of representations—cannot be exorcised from the cognitive sciences. (shrink)

In this paper, I argue for phenomenology, Husserlian phenomenology to be precise, as providing a solid paradigm on how to determine and assess hallucination. To be more explicit, in the context of my deliberations, I analyze Susanna Schellenberg’s arguments for “phenomenal” evidence and “factive” evidence, as regards her evidential theory of perception. To pinpoint the inadequacies raised in her account of (the hallucinating) Hallie and (the veridically perceiving) Percy sharing any kind of evidence, I propose Edmund Husserl’s epistemic fulfillment (...) as a detailed epistemological analysis of perception, which takes into consideration the latter’s phenomenological complexity. (shrink)