In a prior recent discussion on free will, I had proposed an “experimental” instead of purely philosophical or theoretical notion of free will to replace the less well defined ones more commonly used. My definition was that a being possessed free will, if:
- It was under no measureable constraints forcing its choices.
- Its results of choices could not be predicted, and
- It “weakly” satisfied the Turing test, in that it could explain a rational for its choice.
In the following discussion two main objections arose, one in which questions whether this excludes things we would normally not think have free will, say AI programs running on a PC or PCs and whether deterministic beings, i.e., beings whose function is driven by deterministic physical laws, can have free will. By my definitions, the first is excluded in that a “listing” of the program and careful examination of code, data, and inputs of the program should be able to allow one to violate #2, i.e., one could predict the response. In that way, the “deterministic” nature of the being means it is constrained too highly by the simplicity (?) of its design to be free.
Determinism however in and of itself does not constrain behavior of a dynamical system as highly as might be expected. Any non-trivial many body dynamical system exhibits a complicated enough phase space that it is in practice unpredictable. Add non-linear interactions and the phase space likely becomes dense (not bound to rational solutions) … I have no proof at hand for that, it’s more something of a conjecture but I think it is right. What is meant by this? Take a simple two body system with an inverse square law interaction. The solution is linear and solveable. Given initial conditions the positions and velocities of the two bodies will be known to be constrained to be found on particular elliptical curve. The error of your prediction will be bound by an error function which is polynomial and likely linear. Take a slightly more complicated, still linear, system and the much higher dimensional phase space will more likely be, although rational, a space filling curve bound by the energy of the initial system. The error of prediction will not be such a kind function as the linear case. Thus given a random time ahead in the future, one will need infinite accuracy specifying the initial conditions to establish the outcome.
Take a even more complicated system, such as the human brain, with billions of interaction and non-linear terms involved and all bets are off regarding any hope of prediction. A simple non-linear one variable iterative formula at certain tuning values can act as a simple “shift” function, stripping off with each iteration the highest order digit. There is no hope of specifing anything with complete certainty, even without resorting to quantum mechanics which tells us it is really impossible, so non-linearity as in the weather defeats prediction without resorting to the quantum randomness inherent at small scales. And it really is a different sort of randomness I think. Quantum randomness needs short distances, high energies, to tap. Classical chaos gets its randomness by a different road.
One question that comes to (my) mind is the following. It seems to me that a source of randomness, quantum or deterministic, can alone in concert with system which can yeild aesthetic judgement might yeild creative results. Let me restate that. If one could develop a deterministic (rule based for instance) algorithm, network, or other system which can evaluate an aesthetic, i.e., judge a thing on its artistic worth or for lack of another term … judge beauty. Can you develop a system to judge by a set criteria, whether music played or a score is beautiful? Because if that were the case, then determinism might be compatible with creative genius. A finite pool of monkeys will not rewrite Hamlet (or the unwritten unthought sequel) unless the monkeys (or the judge) can filter the good from the bad. But if you learn to see beauty in a system, like music or maths, and tap into randomness in a productive way, it can drive new thoughts onto the board unconsidered by anyone, but which pass the aesthetic judge. And from thence, one might get the little Mozart-like comic operatic ditty transformed, as Shostokovitch did in his magnificent 7th, into a multi-faceted representation of the Nazi war machine.
On the “simple” system and it’s exclusion, Anne had written:
Here’s what fails with the “like a duck” definition: there are computer programs now (e.g. story-generators) that, from the outside, walk like a duck and talk like a duck, and when confronted with those you want to get inside the box and know initial states and step-by-step debugging: you want access to information you cannot have for a human being.
My answer is the following. The human is a fundamentally different dynamical system. You cannot get inside the box, know its initial states and do step by step debugging. The computer running the AI program is akin, in a fundamental way, to the two body problem. It is completely defined and constrained by its initial conditions. The human is too comples, likely non-linear and chaotic. It’s “inital states” cannot, fundamentally, be known. They are a source, not of pseudo-randomness, but real actuall randomness.
Creativity, to my mind, cannot exist without free will. If one is faced with a choice or a problem a constrained system, i.e., unfree, will be constrained to choose a result pre-ordained. What a constrained being cannot do, is think or create ideas which are new, unconsidered by the constraint. A will must I think, by defnition be free to think a thought unthought by anyone before.
She also asks a kicker of a question:
But lack of knowledge (of constraints) is not the same as knowledge of lack of constraints, like with the limited knowledge of the prisoner in the two guards dilemma. 9 times out of 10 you’re an ontologist not a functionalist; why are you a functionalist now?
Ontology is a new game for me. 😀 You (Ann) was trained in phychology and have been a programmer for 2 decades. I was trained in Physics and have also been a progammer for 2 decades. The functional definitions, the positivist approach to theory and experiment was hammered in for a decade before I knew ontology was a word that meant anything. It is an interesting question of why JA and I have traded hats though, why has he donned the ontological hat and I the positivistic one? But of course the more interesting question might be is whether the positivistic definition of free will connects at all with an ontic one.