Pedro Mediano kindly agreed to do a Q and A session with us, particularly based on two of his papers:

The full discussion is available here

Some highlights from the discussion

These answers are paraphrased for convenience - see the video for full responses. We tried not to take anything out of context though!

We didn’t find the distinction between selfhood and the sense of self super clear, and how each of these relate to causal decoupling

  • selfhood is the lower level property - more associated with a Markov blanket - being demarcated from the environment as an individual
  • sense of self is more the phenomenological aspect - the feeling of persisting over time

What would non-emergent consciousness feel like, given that $\Psi$-ID says it’s possible?

  • This relates to the early versions of IIT, where transfer-only systems can have some $\Phi$, so should have some consciousness
  • Most “interesting” systems aren’t of the transfer-only type though, so they haven’t been examined in too much detail
  • So, we can’t really say what it would feel like to be this system

Could we say transfer atoms relate to access consciousness and synergistic atoms relate to phenomenological consciousness?

  • This is an interesting idea, and testable by a conscious access task - whether changes in the task are associated with changes in transfer but not synergistic atoms

As a follow up, how can we assess phenomenological consciousness without also assessing access consciousness?

  • No report paradigms (eg. using pupil size) are the most promising avenue of investigating phenomenological consciousness

How do we decide which atoms contribute to consciousness? Is this likely to change in the future?

  • We should move away from a hard dichotomy of particular atoms either contributing or not to consciousness.
  • We should more think that particular atoms are associated with particular cognitive phenomena, some of which we typically group together and call “consciousness”

What is the phenomenological effect of psychedelics according to $\Phi$-ID?

  • They lead to less causal decoupling (integration), which leads to a reduced sense of self - ie. ego dissolution
  • Experimental work has seen causal decoupling being reduced in the default mode network with psychedelics, which correlates with experimental reports of ego dissolution
    • These findings will hopefully be published soon

Is there some danger of strong and weak IIT drifting apart? How do we make sure they stay together and continue inform each other?

  • The core principles of both are similar, as are the reasons for approaching either of these theories - so hopefully they continue to inform each other
  • A hypothetical example for how this might work - we investigate the transfer atom and see that it doesn’t have an effect on consciousness in a wide ranges of conditions. It could then make sense to update strong IIT so that transfer-only systems don’t have $\Phi$

Global workspace says that consciousness is at the front of the brain. IIT says it’s at the back in the posterior hot zone. How can synergistic global workspace combine these two contradictory predictions?

  • Actually, $\Phi$-ID predicts that the default mode network (in the back) and the frontal parietal network (in the front) both contribute to consciousness, but in different ways - as seen through the decomposition
    • Hypothesis - the sense of self is mostly in the back and metacognitive aspects and access consciousness are more in the front
  • Experimental evidence points more towards the back, but IIT’s theoretical arguments for why this is the case aren’t particularly strong

What about the explanatory gap between the real and hard problems?

  • Like Anil Seth, Pedro thinks solving the real problem will dissolve the hard problem, so it’s not something we should concern ourselves with necessarily

Rank the IIT axioms

  • Overall, “all the axioms need work”. A good reference for this viewpoint is here
  • Integration and differentiation aren’t particularly controversial [given they show up in weak IIT, this isn’t a huge surprise], but the others not so much

What are the ethical implications of a measurable theory of consciousness? Eg. for a “consciousness detector” - trading off false positives vs false negatives

  • Better understanding consciousness should lead to new ethical developments
  • This viewpoint is explained more here