Estimate the Substrate of Consciousness from the ‘Connectome’ of the Human Brain
INSTITUTE: University of Wisconsin-Madison / Department of Psychiatry / Center for Sleep and Consciousness
PROJECT LEADER: Giulio Tononi
PROJECT DURATION: 3 years
LAB WEBSITE: Center for Sleep and Consciousness
Can we establish, from first principles, which regions of the brain can support consciousness, and if so, why? Can we then validate these theoretical predictions against empirical evidence from our own consciousness? Doing so is a crucial requirement for a scientific understanding of what makes us who we are, what our limits are, and how they can be modified and expanded. It is also the only starting point from which we can hope to meaningfully extrapolate to other systems—natural and artificial—and be able to assess if there is ‘something it is like to be’ them.
Integrated information theory (IIT) is unique because, unlike other scientific approaches to consciousness, it starts not from neuroscience, but from experience itself (Tononi et al., 2016). IIT identifies the essential properties that are immediate, indubitable, and true of every conceivable experience: every experience is subjective, from the intrinsic perspective of the subject of experience (intrinsicality); structured, being composed of phenomenal distinctions and relations (composition); specific rather than generic (information); unitary, rather than reducible to non-interacting components (integration); and definite, having a border and a grain (exclusion). IIT then translates these properties in physical terms and develops measures to determine whether a physical substrate can support consciousness, how much (as quantified by integrated information (Φ), and of which kind.
In this project, we will use high-quality fMRI datasets obtained by the Human Connectome Project to develop software that can estimate Φ based on systems’ connectivity. For this purpose, we will leverage high-quality, massive datasets obtained by the Human Connectome Project (Glasser et al, 2016) and will analyze functional connectivity matrices according to key postulates of IIT, namely intrinsicality and information (using estimates of mutual information between current and future activity within a set of brain regions); composition (by estimating the φ value for subsets of the matrix); integration (by employing a divergence measure between intact and partitioned subsets); and exclusion (by approximating maxima of integrated information over the functional connectivity matrix).
We will then be able to test the prediction from first principles that the full NCC, as characterized in a companion Project (Establish which brain regions constitute the physical substrate of consciousness in the brain), has an underlying architecture conducive to high Φ, whereas other brain areas do not. Preliminary analyses suggest that posterior cortical areas have functional connectivity properties supporting high Φ, and these areas are the current best empirical candidate for the substrate of human consciousness, By contrast, brain networks whose architecture is not conducive to high Φ, such as the cerebellum and parts of prefrontal cortex, do not seem to contribute directly to the content of consciousness.
If these predictions are correct, we will also evaluate a principled explanation for these findings. According to IIT, the ability of different brain regions to contribute or not to consciousness depends on certain graph-theoretical properties of their anatomical connectivity, which determines their ability to support high Φ.
Broader Impact:
This Project will test whether it is possible to predict the neural substrate of our consciousness purely based on its essential phenomenal properties—those that are true of every experience. Based on integrated information theory (IIT), these properties can be translated in physical terms. In particular, IIT predicts that the neural substrate of consciousness should be a maximum of integrated information (Φ). If the results indicate that this maximum fist with the empirical evidence about the neural correlates of consciousness, we will be able to make principled inferences about the presence or absence of consciousness in other systems, including species with different brain as well as computers (see companion Project: Estimate integrated information for computer architectures and demonstrate a dissociation between artificial intelligence and consciousness).
Publications:
Barbosa LS, et al. "Mechanism Integrated Information", Entropy, 23.3, (2021): 362.
Grasso, Matteo, et al. "Causal reductionism and causal structures." Nature neuroscience, 24.10 (2021): 1348-1355.
Ellia, Francesco, et al. "Consciousness and the fallacy of misplaced objectivity." Neuroscience of Consciousness, 2021.2 (2021): niab032.
Gomez JD, Mayner WGP, Beheler-Amass M, Tononi G, Albantakis L. Computing Integrated Information (Φ) in Discrete Dynamical Systems with Multi-Valued Elements. Entropy, 23.1 (2020): 6.
Barbosa, Leonardo S., et al. "A measure for intrinsic information." Scientific reports 10.1 (2020): 1-9.
Mensen, Armand, William Marshall, and Giulio Tononi. "EEG Differentiation Analysis and Stimulus Set Meaningfulness." Frontiers in Psychology 8.1748 (2017).