Suppose you built robots. Suppose also that you had computers of your own, immeasurably more powerful than those that you built into the robots. Now suppose that a feature of the the learning capability of the robots was that, when they failed in any tasks, they responded by randomising the values of the inputs used in deciding how to perform those tasks. The more frequently they failed, the further up the logical tree of assumptions involved in deciding how to perform those tasks, or groups of similar tasks, they randomised the values of the imputs. This would be quite a logical response, as if they were saying to themselves, “well, if I failed in that task, it must be because one of my inputs was incorrect; and since I have no idea why it was incorrect, I should experiment by randomising the value for that input and see whether my success rate improves or not.”
I have thought for a long time that free will is basically a process of controlled randomisation; certainly scrutinising the way I use my own free will indicates to me that what I do is simply try things at random when I feel frustrated by a situation, rather than repeat the same rote response indefinitely.
Whether this is an adequate model of free will or not, it gives me some idea of how God relates to humans and other free-willed beings. His knowledge is, as I put it, immeasurably greater than ours, because He is not time- or space-bound, but present at every moment in time and every point in space simultaneously. Therefore He can, as it were ‘anticipate’ every decision, even the random ones, and alter the circumstances at what we would regard as being earlier points in time in order to structure the way our free will will operate at what we would regard as later points.
Note that this model is neutral with respect to the question of consciousness: it makes no difference to the model whether the robot is conscious or not (whatever that means).
