IIT is a proposal to explicate the concept of consciousness by information theory. More precisely, by a mathematical formalization and a prescript how to determine the degree of consciousness of an information system.
In a survey article, "Reggia, JA: The rise of machine consciousness: studying consciousness with computational models." (https://www.ncbi.nlm.nih.gov/pubmed/23597599) compares four current approaches to study consciousness with a computational model. In characterizing IIT he writes:
The basic idea is that the shared or mutual information among brain
regions that is above and beyond their information content as
individual regions enables them to interact in a constructive fashion.
This information integration is viewed as a neural correlate of
Like for any scientific theory, falsification or confirmation of the theory depends on comparing its consequences with observable facts. Here Regia comments on IIT:
On the practical side, a major barrier to advancement in this area is
that one cannot apply φ directly to neural networks of appreciable
size. Further, a φ value alone does not really provide a meaningful
indication of whether consciousness is present or not: without a
golden standard such as the φ value for a typical human brain to use
as a comparison, it is extremely difficult to interpret what any given
value of φ actually means (Gamez, 2010). Such a measurement for real
brains is currently out of reach, not only because of the size of the
networks involved but also because our inability to precisely
characterize the detailed neural circuitry of the brain.
I can send you Reggia's full paper on request.