Last year's FQXi conference was a memorable event for me. Not only because it doesn't happen all that often that my conversation partners abandon their arguments to hurry away and find some more of these pills against motion sickness. But also because I was reassured that I am not the only physicist with an interest in questions that might become relevant only so far into the future that the time spent on them rests precariously on the edge between philosophy and insanity.
Scott Aaronson (who blogs at Shtetl-Optimized) went ahead and gave a talk about free will (summarized by George Musser here), which in return encouraged me to write up my thoughts on the topic, though I've still hidden them in physics.hist-ph (which I previously didn't even know exists!):
So, here's the executive summary.
In my previous post on free will, I explained that I don't buy the explanation that in a deterministic universe, or one with a random element, free will exists as an emergent property. If you call something free will that emerges from the fundamentally deterministic or probabilistic laws that we know, then I can't prevent you from doing that, but there isn't anything free to your will anymore. If free will is as real as a baseball, then you have as much freedom in making decisions as a baseball has to change its trajectory in Newtonian mechanics, namely none.
You might seek comfort in the fact that it is quite plausible that nobody can predict what you are doing, but this isn't freedom, it's just that nobody is able to document your absence of freedom. If your "will" is a property of a system that emerges from some microscopic laws, its laws might be for all practical purposes unknown, but in principle they still exist. If time evolution is deterministic, any choice that you make now strictly followed from an earlier state of the universe. If time evolution has a probabilistic element, as in quantum mechanics, then choices that you make now must not necessarily follow from earlier times, but you didn't have any choice either because the non-deterministic ingredient was just random.
Needless to say, I have greatly simplified. Notably, I've omitted everything about consciousness and the human brain. Look, I'm a particle physicist, not a neurologist. The exact working of the brain and the question whether quantum mechanics is relevant for biological processes don't change anything about the actual root of the problem: There is no room for anything or anybody to make a decisions in the fundamental laws of Nature that we know.
One way out of this problem is to believe in what is known as "strong emergence." That would be if the laws of the macroscopic emergent systems (e.g. "you") do not follow from the microscopic laws. The only people I have met who managed to make sense of this idea are philosophers. There is presently no formal way to achieve such a behavior and there is no known example how this could work. (We discussed here a paper that made an attempt into this direction, but note that the assumption of an infinite rather than a large system is crucial for that to work.) But yes, finding an example for strong emergence would be a possibility. Just that I couldn't find one.
My paper is much simpler than that. In my paper I just pointed out that there exist time evolutions that are neither deterministic nor probabilistic, certainly not in practice but also not in principle. Functions that do that for you are just functions physicists don't normally deal with. The functions that we normally use are solutions to differential equations. They can be forward-evolved or they can't and that is exactly the problem. Yet, there are lots of functions which don't fall in this category. These are functions can can be forward evolved, yet you have no way to ever find out how. They are deterministic, yet you cannot determine them.
Take for example a function that spits out one digit of the number π every second, but you don't know when it started or when it will end. You can record as much output from that function as you want, you'll never be able to tell what number you get in the next second: π is a transcendental number; every string that you record, no matter how long, will keep reappearing. If you don't know that the number is π you won't even be able to find out what number the algorithm is producing.
The algorithm is well-defined and it spits out numbers in a non-random fashion that, if you'd know the algorithm, is perfectly determined. But even if somebody monitors all output for an arbitrarily long amount of time to an arbitrarily good precision, it remains impossible to predict what the next output will be. This has nothing to do with chaos, where it's the practical impossibility of measuring to arbitrary precision that spoils predictability: Chaos is still deterministic. The same initial conditions will always give the same result, you just won't be able to know them well enough to tell. Chaos too doesn't allow you to make a choice, it just prevents you from knowing.
But what if you'd make your decisions after a function like the one I described? Then your decisions would not be random, but they wouldn't be determined by the state of the universe at any earlier time either (nor at any later time for that matter). You need to have your function to complete the time evolution, which is why I call it the "Free Will Function."
This is far too vague an idea to be very plausible, but I think it is interesting enough to contemplate. If you would like to believe in free will, yet your physics training has so far prevented you from making sense of it, maybe this will work for you!
If you found this interesting, George Musser has storified the topic, so you can continue reading.