I've drawn a simple figure to try and capture what's on my mind:

- MI & action learning 2.JPG (48.68 KiB) Viewed 83 times
In the figure I have reformulated the control model as the four phases of the action learning cycle. The different coloured lines represent the potential "intelligences" that could be operating during a particular phase.
When people think about learning/optimisation, it's usually unidimensional. For example, my plan is to jump over the fence; I've never done it before so my plan of action is to wait and see how somebody else manages it before I try it myself; I observe somebody jumping over the fence; I evaluate those observations against my capabilities and my original objective; I now have a model of how I could do it; I put it into practice etc etc. With a unidimensional approach, one would assume that the principal intelligence involved here is kinaesthetic. If one were to follow it through on the figure, then only one of those lines would light up.
However, it's obvious that our cognitive/behavioural processes are more multidimensional, where any process of optimisation/learning involves more than one of those 'multiple intelligence' lines lighting up. But this could happen in a whole series of different patterns: lines lighting up in parallel; different lines becoming dominant at different stages of the optimisation/learning cycle; some lines remain lit up at a particular stage while other lines progress to different stages; etc.
I think we probably have to explore this by looking at the literature in cognitive/behavioural science. For example, I was listening to a series on BBC Radio Four looking at the evolution of physical objects for human use, and at one point the presenter mentioned some cognitive science research that showed the same area of the brain lighting up when fashioning a flint knife and when speaking.
What I want to get at is to move beyond the rational/intuitive dichotomy in thinking/acting. In other words, develop a more comprehensive model of cognitive engagement with complexity.