arXiv Preprint Archive
September 5, 2025
Shahar Dror, Dafna Bergerbest, Moti Salti
The encounter between artificial intelligence and consciousness research is often seen as a challenge to determine whether AI systems could be conscious, but it also offers an opportunity to test and expand existing theories. Current approaches are polarized: computational functionalism focuses on abstract organization and neural correlates, while biological naturalism ties consciousness to living embodiment. Both risk anthropocentrism and limit recognition of non-biological subjectivity. To move beyond this impasse, the authors propose a dual-resolution framework combining the Information Theory of Individuality, which defines ontological conditions of informational autonomy and self-maintenance, with the Moment-to-Moment theory, which specifies epistemic conditions of temporal updating and phenomenological unfolding. This integration reframes consciousness as the epistemic expression of individuated systems in substrate-independent terms, offering a generalizable theory and positioning AI as a testbed.
Frontiers in systems neuroscience
January 1, 2022
Daniel Revach, Moti Salti
Understanding the mind and its relation to the body depends on how consciousness is defined. The authors argue that an information-theory perspective offers a unified and parsimonious account. They present consciousness as the propagation of information from the past to the future, contrasting with dominant models that treat it as a discrete brain function. Using this lens, they demonstrate that temporal continuity in perception can explain the emergence of volition, subjectivity, higher-order thoughts, and body boundaries without additional assumptions. The work discusses implications for the mind-body question and the appeal of embodied cognition.
Consciousness and cognition
November 1, 2021
Daniel Revach, Moti Salti
The way researchers ask questions about consciousness is shaped by hidden assumptions that constrain possible answers. This paper identifies and tests the assumptions underlying contemporary debates about the neural correlates of consciousness, tracing them to Crick and Koch's 1990 framing. These premises create a sequential, passive view of conscious perception as the product of resolved unconscious processing at a single moment in time. Major debates—about the moment of entry, all-or-none versus graded nature, and report versus no-report paradigms—are driven by consensus on these assumptions. Removing these assumptions resolves some debates and suggests a dynamic, dispersed conception of consciousness, offering a more parsimonious account of conscious perception.
eLife
May 5, 2021
Moti Salti, Simo Monto, Lucie Charles et al.
Conscious perception of a stimulus is linked to a specific sequence of brain processes that unfold after about 270 milliseconds, involving superior parietal and superior frontal regions. Using EEG and MEG recordings during a task where participants reported the location and visibility of a brief visual target, multivariate pattern analysis showed that information about stimulus location is still processed in the cortex even when the stimulus is not consciously seen (blindsight). However, only consciously perceived stimuli trigger additional, time-limited neural activity in higher-order brain areas, whereas unconscious processing decays slowly without this chain of activity.
arXiv Preprint Archive
March 24, 2018
Moti Salti, Asaf Harel, Sebastien Marti
Conscious perception may not be a discrete event but a dynamic process in which a stimulus is continuously recoded to fit an ongoing stream of other perceived stimuli. This hypothesis can reconcile inconsistent findings on the timing of neural correlates of consciousness and makes testable predictions. The updating process is governed by context, stimulus saliency, and the observer's goal. The framework challenges the typical distinction between conscious and unconscious information processing.