Daniel C. Dennett
Quite interesting, and introduces more stuff that is new to me, some in areas I'm interested in.
I think that Dennet's critique of the Libet experiments does show that it's hard to assign any particular interpretation to them, especially with respect to the apparent placement of conscious decision after the action, due to the unclear timing relationships of the visual perception of the “stopwatch” display. What I find more compelling is the “action potential” which precedes the action by 600ms or more. I.M.O. This is likely an unconscious precursor to the conscious action, but I also note that the structure of the experiment may cause particular results, eg. In setting it up so that there is no particular reason to act at any time, so it comes down to intuitively judging that it is “about time” to do something, and there must necessarily be a slow unconscious process underlying this. Dennet notes that other similar experiments give significantly different results, for example asking to push immediately after a flash is much faster (250ms.) It seems that we can set up a “reflex” connection for this sort of task, but if asked to delay the response the minimum possible amount, they end up delaying an additional 300ms, seemingly because some different mechanism is involved (I suspect that people with musical training could do better.)
It's virtually tautological that if the process of “being conscious” or “consciously deciding” takes any time, and if we can somehow sense the beginning of this process, then we will sense an unconscious precursor to the consciousness event. This should be unsurprising, but I found it fascinating because it seemed to show that the unconscious drove the conscious. Once you think of consciousness as taking time this is obvious, but one tends not to think of consciousness as taking time, so it is still interesting.
He mentions the possibility of the “out of the loop” consciousness, but clearly doesn't prefer it, and feels that the experiments where people confabulate reasons are artificial and don't represent normal operation. I've been favoring the “out of the loop” idea, and I also have an architectural explanation of why we should expect consciousness to be out of the loop, which is that subconscious processing is probably largely in the form of fuzzy neural net processes which don't operate symbolically, so it is impossible for them to report out an explanation. This same issue is seen in AI. It's just an oracle, and we are faced with coming up with an explanation based on the “confirmation bias” information that it throws up into consciousness.