Dear Turil,
you ask whether I have evidence that humans can flip themselves, in the sense that they can modify their behavioural rules, while cellular automata can`t.
I suppose this can be seen as an instance of the tough question whether we are completely determined in our behaviour by some fixed rule, or we enjoy free will. I believe there exists a scenario in which both things are in some sense true. This is made possible by the fact that this universe is multilevel.
At the bottom level - the ultimate spacetime scale from which everything emerges computationally, including humans - rules are algorithmic, and fixed; or, if they change, this is not under our control.
But at the much higher level of our direct experiences - say, the biosphere - we feel we are able to change our own behaviour. This is illusional, since, under this scenario, even the fact that I have decided to behave better, or worse, is determined by the rules and dynamics at the lower levels, and yet the illusion works fine for us, since we cannot directly experience the causal influences of those lower dynamics on what happens to us.
I also have in mind the argument used by Wolfram for preserving a (weak) notion of free will in a fully deterministic, computational universe. His idea is that it takes no less than 10 computational steps, in a simulation, to find out how the universe, or my life, will look like in 10 universe steps from now, since the computation that the universe is performing is irreducible - no shortcut. Thus, the rule at the bottom is fixed (no free will), but the emergent behaviour appears unpredictable, thus, in a sense, rule-free, spontaneous.
If this is a bit confusing, don`t worry. It is for me too. Trying to capture the notions of spontaneity, agency, creativity, free will, in a formal framework, is always troublesome. It is probably the hardest problem, when trying to formalise Teilhard`s cosmological views.