1 - Integrated Information Theory

Hi Jason! This is a kind of side note to my ASC work, but I thought it was super interesting and wanted to ask what your thoughts are (I know I’ve written quite a bit, but I wanted to try and explain myself properly).

I’m not sure if you remember, but a while back we had a discussion on the wiki r.e. the idea of consciousness… Catalyst - Custom Universe.

Yesterday I went to a very interesting seminar presentation by Giulio Tononi entitled ‘Consciousness and Cause-Effect Power’. (Funnily enough, the presentation was introduced by David Chalmers, but I only realised this just after I had left!)

Anyhow, the presentation introduced Integrated Information Theory (IIT) - a framework Giulio had come up with to quantify consciousness. I’m not sure whether you know much about it - I got the impression that it was quite a recent theory - but the main ideas seemed to be the following:

  1. It is not possible to define consciousness by starting with physical matter and trying to figure out what this means for the mind. So the only way forward is to start with the mind and determine what this means for a physical system (i.e. define consciousness in terms of ‘experience’, and then figure out how this would need to be implemented in a physical system).

  2. He then introduced five axioms, or ‘criteria’, for conscious experience; existence, composition, information, integration and exclusion. He explained each of these in a fair amount of detail. (Exclusion was the one that interested me the most - basically it was the idea that two ‘brains’ cannot ‘overlap’ in terms of consciousness. I thought this addressed some of the questions we had been discussing on the wiki.)

  3. He then defined consciousness via an idea called ‘cause-effect power’, i.e. a conscious system is one that exerts a direct effect on itself, independently of its external environment. (The system modifies itself, by itself, and this effect is independent of ‘sensory’ input/ output to the physical system.)

  4. This was where it got a bit complicated. I didn’t understand the intricacies of this but he took these ideas and explained how he had used these to ‘map’ consciousness in terms of networks of cause-effect power. It all looked very mathematical and a bit beyond me. From here he was able to quantify the spectrum of consciousness in a range of cases. A simple and/ or loop, for example, was conscious according to IIT! (albeit so minimally that it wouldn’t really mean much)

There were certainly a few people in the room who had reserves about the theory (there was an particularly interesting discussion about its disagreement with AI), but I don’t actually think anyone was able to raise any damning objections to it in the end.

Giulio did also note that IIT is a scientific theory, in the sense that further experimental work is required to support it. But that nevertheless it is a good starting-point for quantifying and predicting consciousness in physical systems.

A couple of questions:

  1. Basically, I’m just curious as to whether you think this idea has any merit. I have to say, at first I was sceptical, but Giulio responded so thoroughly and convincingly to every objection that he really started to convince me (at least that it’s a good starting point for a definition).

  2. In terms of the ASC, I also wanted to clarify the difference between ‘sentience’ and ‘consciousness’. Whilst an and/or loop might be considered conscious (according to IIT), I’m not sure it would be considered sentient…at least according to the definitions I’ve come across. Are these two separate ideas? I guess figuring out exactly what sentience is is a good place to start. —– Just then I re-read your notes from our phone conversation and realised you mentioned exactly this question! If you could point me in the direction of some Philosophy of Mind stuff about this (as you suggested) that would be great :)

Jo

Well done for finding something so interesting so quickly!

I have a general expectation that anyone trying to bite off so much in a brand new theory is likely to be wrong. Also information theory in general is trendy in just the wrong way at the moment, IMO. But let’s give this person a fair go! So …

  • I like the idea of starting with experience. IMO, experience, consciousness and sentience are all the same thing, and experience is my favourite of the three words because I think it has a very clear meaning in everyday language.

  • “It is not possible to define consciousness by starting with physical matter” — This seems obviously true to me, at LEAST given the state of current science even if it’s not true in principle. (If in doubt about this, see http://bunny.xeny.net/temp/Jo/What-is-it-like-to-be-a-bat%3F.pdf.))

  • “cause-effect power” sounds to me like it’s COMPLETELY changing the subject away from consciousness and to something else. I guess you don’t agree, and you know this theory better than I do. Is there anything you can say to help me with this particular issue? If you’re not sure at this point, you could do Crash Course In Philosophy Of Mind and then we could come back to Tononi.

  • “A simple and/ or loop, for example, was conscious according to IIT! (albeit so minimally that it wouldn’t really mean much)” — IMO this is not necessarily stupid, as long as it’s only very minimally conscious. And I think it would actually be OK for the ASC, if you ended up believing it. The main problem with it, IMO, is just the one above: it depends on believing that consciousness is related to cause-effect power.

Jason