arXiv (Cornell University)
June 1, 2026
Ulysse Klatzmann, Adrien Doerig
Progress in AI is turning machine consciousness from a philosophical curiosity into a societal issue, prompting criticism of computational functionalism. Biological Naturalism (BN) claims biology, not computation, is crucial for consciousness. Two forms are distinguished: Type-A BN holds that biology intrinsically matters for consciousness without affording unique information processing capabilities, which dissociates consciousness from behavior and makes it untestable. Type-B BN holds that biology matters because it affords unique information processing capabilities; this form is testable and not incompatible with computational functionalism. Both face the same task of relating consciousness to information processing, with biology serving as a guide but not a solution.
Cambridge University Press eBooks
April 23, 2026
Michael H. Herzog, Aaron Schurger, Adrien Doerig
This paper critically examines quantum theories of consciousness, particularly orchestrated objective reduction (OrchOR), which proposes that proto-conscious events occur throughout the universe but full consciousness arises only at the subcellular level in neuronal microtubules. The authors outline challenges with this approach, arguing that existing theories linking consciousness to neural processes are insufficient and that quantum-based explanations face significant conceptual and empirical difficulties.
Philosophy and the Mind Sciences
December 22, 2025
M. Päßler, Adrien Doerig
Conscious experiences are defined by their relationships to one another, and these relationships are mirrored by brain structures. However, simply matching brain patterns to mental patterns is not enough to explain what a conscious experience is about. The brain's downstream processes must actively use those patterns in a way that preserves their structure and influences behavior. Purely anatomical or overly broad causal brain structures fail this test, but activation patterns can succeed when embedded in the right computational context. This means that local structuralist theories, which ignore how brain activity is used by subsequent processing, are incomplete. Any adequate structuralist account of consciousness must incorporate computational context.
The Behavioral and brain sciences
June 25, 2026
Michael H Herzog, Adrien Doerig
Fleming and Michel link consciousness to sensory horizons, model-based computations, reality monitoring, and decision termination, but these functions can operate without consciousness, potentially making consciousness epiphenomenal. Like other current theories, this approach places excessive explanatory weight on a limited set of factors.
Consciousness and Cognition
February 1, 2022
Michael H Herzog, Aaron Schurger, Adrien Doerig
The authors argue that integrated information theory (IIT) cannot be a scientific theory of consciousness because it relies on pure first-person experience as a foundation, which is not measurable except through reports, brain activity, and their relationship. Science requires measurements and data, not assertions of ground truth independent of data. No experiment based on third-person data can test IIT as a theory of consciousness. The authors contend that IIT's conception of consciousness is fully dissociated from measurable things, implying that both the level and content of consciousness are epiphenomenal and lack causal power. This leads to a form of dissociative epiphenomenalism that undermines trust in reports about first-person experiences, which are the basis for IIT's own axioms, meaning accepting IIT leads to rejecting its axioms.
Cognitive Neuroscience
July 14, 2020
Adrien Doerig, Aaron Schurger, M. Herzog
Consciousness research has produced many competing theories, ranging from computational to quantum approaches, more than in other natural sciences. This abundance may stem from a lack of clear criteria for how empirical data should constrain such theories. The authors argue consciousness is empirically well-defined and propose a checklist of criteria that empirical theories of consciousness must address. They review 13 influential theories against these criteria, revealing their relative strengths and weaknesses from a strictly empirical perspective.
Consciousness and Cognition
July 1, 2019
Adrien Doerig, Aaron Schurger, K. Hess et al.
Theories that identify consciousness with specific causal structures in the brain, such as those requiring feedback loops, are either false or unscientific. Using theorems from computation theory, the authors demonstrate that causal structure theories—including Information Integration Theory (IIT) and Recurrent Processing Theory (RPT)—cannot be empirically tested because any causal structure can be realized by systems that lack consciousness. This undermines the claim that feedforward systems are never conscious and feedback systems always are. The argument suggests that consciousness research should focus instead on functional explanations, such as global workspace or higher-order theories, which are compatible with diverse neural implementations.
Neuroscience of Consciousness
January 1, 2019
Adrien Doerig, Frank Scharnowski, Michael H Herzog
The longstanding debate over whether conscious perception is continuous or discrete is addressed with a two-stage model in which unconscious processing integrates information continuously until a discrete conscious percept forms. In response to criticism from Fekete et al., the authors argue that simple sliding windows cannot account for apparent motion and related phenomena within a continuous framework. They also contend that the supervenience argument raised by their critics applies only to qualia realists, a philosophical position the authors do not adopt.
Michael H Herzog, Aaron Schurger, Adrien Doerig
preprint
The authors argue that integrated information theory (IIT) cannot be a scientific theory of consciousness because it relies on pure first-person experience as ground truth, which is not measurable except through reports, brain activity, and their relationship. They contend that science requires testable data, and no experiment based on third-person data can test IIT as a theory of consciousness. The authors further claim that IIT implies consciousness is epiphenomenal—lacking causal power—and that accepting IIT leads to rejecting its own axioms, since reports about first-person experiences, which are the foundation for those axioms, become untrustworthy under the theory.