What I've been thinking about...
“The Secrecy is Essential to the Freedom of the Agent…”
Awhile back, I posted an audio clip of Daniel Dennett talking about chess programs, and how it makes sense to say that one “could have” beat the other, even though they’re deterministic programs.
It was a ten minute clip taken out of the full podcast, which was one hour. On listening again, I may have lost some of the clarity of the connection to the topic of Free Will in the interests of brevity. Because I think there is a worthwhile point here, I’m going to take a stab at a written summary.
If you want to find out how chess programs, Secrecy, and Free Will are looped together into one super-meme, click on the read more link…
If I had to guess, I’d say Dennett believes in physical Determinism. Still, he speaks to justify and define Free Will for the full hour in this talk. Whatever inconsistencies may exist, there are sentences that use the concept of Free Will that really mean something. For example:
Unless they’re being difficult (see picture below) anyone, including determinists, will admit that these sentences have meaning and are true. In his talk, I think Dennett’s aim is to find the definition for a Free Will that is philosophical, but that also applies to common uses like those above.
For Dennett, Free Will is tied up with the idea that something “could be” different than it actually is. Even if I sit with my hands in my lap, I’m freer than I would be with bound hands, because I “could” make any gesture I like.
For a physical determinist, this is a tricky issue. An argument could be made that the deterministic conditions leading to my choice to keep my hands still is just as much of a cage as a rope would be. However, Dennett defines what “could be” means in a way that’s compatable with Determinism.
He says that every “could be” statement has an implicit list of mutable conditions. For example, to say that I “could make any gesture” might really mean that I “could make any gesture in a hypothetical world where my thought process was different.”
We must not take this too far. If *any* hypothetical world is allowed, then it would follow that absolutely anything “could be” true. For example, “I could have leapt into the sky and flown to work today” is false, but if we allow hypothetical worlds where Gravity is different, where humans have wings, etc, we may be tempted to call it true. So we must not allow *all* conditions to be mutable, only some.
How do we choose? Which conditions should be mutable? It seems there are an array of answers “right” enough to pass in everyday English. However, Dennett suggests that conditions that are Secret are the ones we should allow to be mutable.
For example, imagine that Tiger Woods and I both attempted to make a hole-in-one, and we both missed.
“I could have made that shot” I say. My claim is false because I would have to consider a very different hypothetical world, in which non-secret conditions (like my being a uniformly unskilled golfer) are different.
“I could have made that shot” says Tiger. His claim is true, because he only needs to consider hypothetical worlds where the eddies of wind, or the arrangement of the blades of grass is different, and these conditions are secrets to us as we swing our clubs.
Of course, every agent has many secret conditions right inside their own brains (secret to themselves even). I suspect that this is what Dennett would say qualifies them as “agents”.
To drive his point home, Dennett explains how his definition of “could be” functions well even in clearly deterministic situations. The deterministic situation he chooses is pitting two chess-playing computer programs against each other. The audio of this part of the talk is available in this post.
To cover himself, Dennett allows the programs to use psuedo random number generators (which are also deterministic by the way). He then suggests the two programs play 1000 games against each other. He points out that if the random number generators are restarted from scratch, and 1000 more games are played, they will be the EXACT SAME GAMES as the first time.
Now, we start asking questions, like “Why did program A win game number 611?” Or, assuming program A won the majority of the games, we can ask “Why did program A win more?” Dennett astutely points out that the answer “because it was determined to win” is no answer at all. The true answer, he says, should be something like “because program A is better designed.”
We must ask ourself if program B *could have* won those games. The answer is yes if B wins in a set of “reasonable” hypothetical worlds. Such as:
On the other hand, some hypothetical worlds are not allowed in the consideration of whether B could have won:
The items in this list are evidently false, wheras changes like those in the first list would be secret. Roughly speaking, saying truthfully whether or not these chess programs could have behaved differently (even though they are deterministic) is consistent with Dennett’s definition of “could have”.