[FRIAM] falsifying the lost opportunity updating mechanism for free will

[∄ uǝlƃ]

To restate it, the mechanism consists of:

• a mesh of parallel processes evolving in time
• each process has a local branching structure for what might happen next
• these branches (and the events that walk them) compose
• that composition is monitored and remembered within some scope
• that monitor/memory is used by a controller to edit the branching structures

What we call "free will" is the extent to which, and perhaps the *shape* of, the branching structure(s) change over time. It's infeasible to measure the branching structures directly, especially 10 years later trying to decide if your mom's an alcoholic or not. But we can estimate the wiggle in the composite behavior over time and retro-infer whatever branching structure monitoring, remembering, and editing might have taken place.

I think to adequately falsify this mechanism, we could implement a few (several would be better) versions of it, sweep their parameters and classify the results. If none of them exhibit clear components and some kind of *sensitivity* in one or more parameters, then the basic mechanism can't generate the phenomena we're looking for.

I think the most important parameters would be the scope of the composer (which processes to include and which to truncate), the fidelity of the monitor, the size of the memory, and the kind of edits (point mutations or something more drastic). It would be validating (and pretty cool) if, say, with a memory size N, entrainment happens quickly, but with memory size N+M, the system flips between 2 behavior/output components. But finding something like that would be a negative result. We'd merely have programmed in the behavior we *wanted* to see come out.

And it would be interesting to include stochasticity peppered throughout to see if that had an effect on the sensitivity or robustness of the output components.

-- 
☣ uǝlƃ