Attention, Neuronal Oscillations and Predictive Coding
Pp. 113-131 (19)
Paulo J. Negro
Theories of consciousness share the idea of consciousness realization in
neuronal states. Conscious experiences take place in embodied agents that generate an
internal model of the external environment and of themselves. Artificial intelligence
researchers have focused their efforts on the creation of a multi-level consciousnessbased
architecture that follows these precepts. Attention mechanisms can serve as
control models on silicon environments with limited resources for information
processing. Conversely, attention can emerge when information-processing modules
cooperate to solve a given problem. However, computationally responding to a topic
does not equal conscious processing. Top-down attention and consciousness are
conflated, but not the same. Phenomenal consciousness seems possible with activation
of local networks, but it is fully realized only after global ignition. Conscious
experiences have their own abstraction level. However, they depend on the dynamic
cross coupling of neuronal oscillations across processing levels. Predictive Coding ties
efferent copies of corollary discharges to the experience of agency. A system based on
monitoring action-based signals may underlie both ownership and awareness. An
efferent copy is an extra copy of a given command from a control region. It informs
local processors of intentions, allowing these downstream processors to timely match
intention with the actual results of motor commands. This voluntary motor model
applies to abstract experiences. A generative model depends on actions. It maps hidden
causes to sensory consequences of these actions. Perception corresponds to the inverse
mapping from sensations to hidden causes. A subject may be inferred from inverse
mapping of the hidden causes that underlie seen or unseen globally integrated actions.
The subject represents the narrative or scenario that predicts sequences of events.
However, caution should be applied against strict localizationism in the interpretation
of this model.
Access Consciousness, Adaptive Resonance Theory, Artificial
Intelligence, Attentional Models, Being the Subject as Inference, Bottom-Up
Awareness, Consciousness-Based Architecture, Corollary Discharge, Cross-
Frequency Coupling, Emotional Tagging, Feelings of Knowing, Generative
Model, Global Workspace, Labeled-Line Limitations, Localizationism, Machine
Consciousness, Neuronal Oscillations Summary, Phenomenal Consciousness,
Primary Visual Cortex, Predictive Coding, Self-Modeling, Systems 1 and 2, Top-
Down Amplification, Variable Q.
Assistant Clinical Professor of Psychiatry and Behavioral Sciences at the George Washington School of Medicine and Health Sciences.