Tutorial on Integrated Information, Causal Density and Conscious Level Part 1
Normal 0 false false false EN-GB X-NONE X-NONE
A key challenge in the neuroscience of consciousness is to develop theoretically grounded and practically applicable quantitative measures sensitive to conscious level. We begin by motivating the hypothesis that, given fundamental phenomenological properties, conscious level must somehow correlate with the extent to which underlying neural dynamics are simultaneously differentiated and integrated. We will describe two groups of proposed measures based on this hypothesis:
(i) Measures of integrated information, which reflect the extent to which the information generated by the whole system exceeds that generated by its parts. We will introduce the concepts and mathematics of information theory necessary to understand integrated information. We focus on two new versions we have developed to overcome previous limitations, which are applicable to realistic neural models and to time-series data.
(ii) Measures of causal density, which characterize the overall causal interactivity between different system elements. Again, we introduce the necessary concepts and mathematics, in this case those of Granger causality which have broad application beyond measures of conscious level.
These measures are gaining prominence in consciousness science, but are more complex than other measures such as synchrony, coherence, etc. There is therefore a need, and likely a strong demand, for a clear account of their formulation and application.
We will contrast and compare the two groups of measures, both in their conception and in their properties in simulation, and discuss – with audience participation – their merits and shortcomings as measures of consciousness. We finish by discussing practical application of these measures to real data.