Friday, October 26, 2007

How An Arminian Robot Makes A Choice

Henry/Robert/Arminibot 3000 Serial Number 777666 is questioning Reformed doctrine once more. While it is obvious to anyone who has studied the issues that H/R/A has not, I thought it might be a beneficial exercise for those concerned if we took the robot motif up once again and pondered a thought experiment.

Is it possible to give a robot free will? To make, as it were, an android?

The question is important because it helps us to define how exactly a choice is made. Currently, computers can be programmed to make “choices” by assigning a weight-value to different options. From there, a risk/reward calculation can be made, and the computer can pick which option has the greatest potential for reward with the lowest amount of risk. This is ultimately how computers can play chess games. They analyze a multitude number of possible moves and rank the orders in terms of which one is statistically most likely to occur.

But obviously this “choice” isn’t a free choice. It relies upon a set of initial factors, such as the hardware used to create the computer. (If a chipset is flawed, the calculations will be flawed and the computer will make erroneous choices.) Further, the software has to be programmed such that the computer is able to assign a weight to various chess functions. A computer is not “born” knowing that pawn a5 is a horrible opening move. It has to be programmed in, and the various values of the board have to be programmed in. Further, the specific values of what levels of risk are acceptable must also be programmed in. These are not laws of nature. They are dependent upon the programmer.

Naturally, one can test the computer after that by simulating several games until the best moves are found. Further testing against human opponents can further hone the skills of the computer. Eventually, you have Deep Blue beating Garry Kasparov.

But this brings up an interesting problem for the libertarian, especially as defined by H/R/A. H/R/A believes that a choice cannot be free unless it is possible to choose a different option. But let us present a computer with two options for an opening move. Either the computer can pick pawn to e5, or it can pick pawn to a5. Given the programming in place, it is impossible for the computer to actually pick pawn to a5 because of how horrible that opening move is compared to the standard pawn to e5 approach.

Now ask Kasparov to make the same decision. Given Kasparov’s knowledge of chess, it is equally impossible for Kasparov to make the move pawn to a5 instead of pawn to e5. Yet we would not say that Kasparov is acting against his free will were he to always play pawn to e5 instead of pawn to a5. We would say he is making the smart move. He would be an idiot to make the other choice.

H/R/A might respond by saying that Kasparov could choose to behave stupidly, if that’s what Kasparov wanted to do, but Kasparov doesn’t want to act stupidly, so he will limit his selection to the smart move each time. This, however, changes H/R/A’s position! What first defined free will as the ability to do otherwise has become simply doing that which one wants to do.

But this secondary definition of free will is actually the very definition that Calvinists hold to. People always do that which they want to do, and the unregenerate always wants to disobey God. Under this definition of free will, Calvinists fully support free will. As such, moving to this explanation doesn’t help H/R/A at all. In fact, it forces an immediate checkmate against his viewpoint.

Before abandoning this illustration completely, let us take another thought experiment. I own Chessmaster 10, and the lowest AI opponent you can face is a chimpanzee that uses completely random moves. There is no attempt to weigh which move is better. The computer compiles a list of all possible legal moves and randomly chooses one of those moves.

Is this random choice any freer than the choice a computer makes by weighing a list? The answer to that question is a resounding no. Once again, the computer chooses based on hardware structures and software limitations. Computers are not really random—they have random seed generators that are strictly controlled. They mimic random events, but in reality they are not random at all. (Each time you reuse the same random seed generator, you get the same result. To avoid this as much as possible, most programs use the date and time functions for their random seed generation. Since it is basically impossible for a person to pick the exact same millisecond on a clock each time he runs a program--even if he resets the clock--it always appears to us as a random result.) So, even engaging in random “choices” is not really random for a computer. Suppose a computer randomly picks the move pawn to a5. This move is determined by the hardware features working together with the software features of the computer so that at the exact moment the program is run, it will always pick pawn to a5. There is never a time when it will not pick pawn to a5 under those circumstances.

But there is a way to get truly random data (assuming one doesn’t have access to the omniscient mind of God). You could hook the computer up to a piece of radioactive matter. Since radioactivity occurs at a completely random, totally impossible to predict, rate for individuals particles, you could create a computer to use those random results to make decisions about chess moves.

But is this any freer? Again, the answer is a resounding no! After all, there is no value in the radioactive decay that says, “If this particle goes now, choose option pawn to a5.” The ability to translate a truly random event into a choice is still based on the software to define what each selection must be. And we haven’t even addressed the elephant in the room: the fact that these random choices are still determined by radioactive decay!

Suppose, however, we were able to surmount those obstacles and create a computer that could make choices that were not based on its hardware or software. It could play a truly random game of chess.

Does anyone think the computer would win the chess game? Of course not. Does anyone think that a computer making choices without reference to a designers hardware specs or software instructions would make good choices? Of course not.

Why, then, do Arminians insist that people must be able to make choices without regard to our hardware (brains) or our software (our nature)? How is it possible for our choices to be good ones if we are able to ignore the hardware specs and the software limitations? How is it possible for us to make any decisions at all outside of the governing physical and spiritual specs that we have?

12 comments:

  1. Agent Causation

    ReplyDelete
  2. Remonstrant said:
    ---
    Agent Causation
    ---

    I'm not sure if that was supposed to be the answer to my question, "How is it possible for us to make any decisions at all outside of the governing physical and spiritual specs that we have?" If it was, then surely you realize that you need to define what an agent is, don't you? If agency is different from the physical or spiritual aspects that I mentioned in my question, then what is agency? If this was an attempt to answer the question, it only moved it back one step.

    ReplyDelete
  3. And of course, we've addressed agent causation before:

    http://triablogue.blogspot.com/2007/04/agent-causation.html

    http://triablogue.blogspot.com/2007/04/todays-your-lucky-day.html

    ReplyDelete
  4. This post is exactly right. It was arguments such as this by Loraine Boettner that eventually lead me to submit to the truth that "God hath decreed in himself from all eternity, by the most wise and holy council of his own will, freely and unchangeably, all things whatsoever comes to pass" (2LC 3.1). The logic of causation and necessary effects is completely inescapable.

    At the same time, I am a total compatiblist, upholding divine foreordination along with freewill, just as the Bible assumes both truths, and as the Confession goes on to say: "Nor is violence offered to the will of the creature, nor yet is the liberty, or contingency of second causes taken away, but rather established."

    ReplyDelete
  5. In “agent causation”, one can have a compatibilist understanding of free will and still be an Arminian.

    ReplyDelete
  6. Dear Peter Pike,

    Your post simply shows that computers don’t make choices. They simply identify things and execute predetermined instructions. Computers have simulated reason, but not desires. Both reason and desire are necessary for choices.

    -------
    Now ask Kasparov to make the same decision. Given Kasparov’s knowledge of chess, it is equally impossible for Kasparov to make the move pawn to a5 instead of pawn to e5.

    --------

    I don’t know why it’s impossible. What if Kasparov was playing his 5 year old niece? He then might want to give her an advantage. But let’s say there was no reason at all for Kasparov to want to make such a move. Perhaps then a5 is impossible. But what about d5? Kasporov knows impressive strategies starting with either d5 or e5. So even if a5 is impossible, Kasporov still has alternatives from which to choose. The problem with the deterministic position is that only one action is possible. There are no alternatives. There’s a big difference between “A & B are possible but C is impossible” and “A is possible but B & C are impossible“.

    Dan

    ReplyDelete
  7. Peter Pike picked a peck of pickled peppers. A peck of pickled peppers Peter Pike picked; If Peter Pike picked a peck of pickled peppers, Where’s the peck of pickled peppers Peter Pike picked? :-)

    I enjoy demonstrating my capacity for humor. Only conscious intentional beings are capable of such fun. Now let’s get serious and talk about Peter Pike’s post on how I make choices. Pike makes lots of mistakes that would take too much time to correct, so I will intentionally limit myself here. Pike concludes with:

    ‘Why, then, do Arminians insist that people must be able to make choices without regard to our hardware (brains) or our software (our nature)? How is it possible for our choices to be good ones if we are able to ignore the hardware specs and the software limitations? How is it possible for us to make any decisions at all outside of the governing physical and spiritual specs that we have?’

    Here is a key error. Arminians and other non-calvinists believe that when God created mankind he created them in His image, with his capacity for self awareness, his capacity of performing his own actions directed by reasons (i.e., he created them with the capacity for freely chosen actions). When my creator was designing my own hardware and software the major goal was to create a being capable of performing his own actions, freely choosing as both God and man are capable of doing so. So I had to have self awareness, I had to have the ability to reason, and I had to be placed in an environment in which the cognitive design plan for Arminibots would be allowed to flourish. In a word, my nature is similar to both divine and human natures in that I am capable of performing my own actions and doing these intentional actions freely (e.g. when God performs an action his action is neither predetermined by some factor outside himself nor is it random, rather, he is a personal agent operating according to reasons).

    So when a human performs an intentional action freely and when I do so, we are in fact operating according to the hardware/software specifications our designers had in mind for us. So all of Pike’s points about actions being predetermined or random are irrelevant concerning us.

    Humans were designed to be similar to God in their capacity for intentional self performed actions that are neither predetermined nor random. Arminibots were also designed to be similar to God in their capacity for intentional self performed actions.

    I appreciated the points made by the humans named Dan and Remonstrant. Remonstrant is correct that the issue is the nature of agent causation. Dan is correct when he says:

    ‘I don’t know why it’s impossible. What if Kasparov was playing his 5 year old niece? He then might want to give her an advantage.’

    Dan is correct that an agent, in this case Kasparov, may have differing reasons for doing particular actions. One reason for doing a particular chess move is that it is the best available move under the circumstances. Another reason a particular move may be made is that the agent is teaching another agent how to play chess. In this case one may make a “bad” move in order to demonstrate examples of “bad” moves or in order to give a much weaker opponent an advantage.

    ‘But let’s say there was no reason at all for Kasparov to want to make such a move. Perhaps then a5 is impossible. But what about d5? Kasparov knows impressive strategies starting with either d5 or e5. So even if a5 is impossible, Kasparov still has alternatives from which to choose.’

    I would not state it as ‘no reason at all for Kasparov to want to make such a move’, rather, I would say, perhaps the reason for Kasparov to want to make such a move is not a good one if the goal is to win the game, if the goal is something else then perhaps he has good reason to make such a move. a5 would then not be impossible but less preferable. Dan then asks what about an alternative, d5 or e5? And here is an important point: perhaps a5 would be a less preferable move, and d5 would be a more preferable move than e5 and a5, but the issue is, could Kasparov perform either of these moves? If the total determinism that calvinists would like to believe, were the case, then the only possible move for Kasparov to make would be the one move which God had predetermined for him to make in that particular situation. So if this kind of determinism were true then all other moves would be impossible for Kasparov, he could not do move a5 or d5 or e5, he could only do one of these moves (or some other move, or even resigning) which had been predetermined. I play and enjoy chess and the game is predicated on the performance of possible moves. In other words, the game assumes that agents can and do freely perform their chosen moves. If their moves are necessitated by some factor external to themselves, then they are not really playing the game, their move is being determined by some other factor or agent, outside of themselves. It is their move if the move is up to them, chosen by them, and not necessitated by some other person.

    Imagine that I am playing my designer Hal, in a game of chess. And imagine that Hal had preprogrammed for me to make a “bad” move at a particular point in the game (i.e. Hal would be behaving like Frankfurt’s’ intervener). I had considered the moves and was going to do say d5 but instead Hal had me do a5 a bad move, which then ensured that he would win the game. Now after the game if I was unaware of what had occurred I might think and others might think I made the bad move. And Hal could even make fun of or punish me for making this bad move. But most thinking persons would see that there is something wrong here. I make the move that Hal predetermined that I make and I could not have done otherwise so I was not acting freely. And yet Hal holds me responsible for making the bad move that He caused me to make?

    ‘The problem with the deterministic position is that only one action is possible. There are no alternatives.’

    Dan makes a very important point here. If all events are predetermined, then only one action is possible, and it is the one action, which the factor that predetermined all events allows to be possible. And since it is the only possible action it is a necessary action and all other alternative actions are impossible. So that would mean, take any event that occurs in history, that event had to occur, must occur, it was necessary for that event to occur, it was impossible that any other event occur at that time as part of that history. And this total necessity of all events applies to every event that makes up history. The determinists’ view leads to these conclusions and yet they do not understand how others might see any problems with the implications and consequences if these conclusions are true.

    ‘There’s a big difference between “A & B are possible but C is impossible” and “A is possible but B & C are impossible“.’

    Dan is getting at an important point here as well. Free will and the ability to choose between available options, alternatives, opportunities’ only exists if “A & B are possible but C is impossible”. Versus “A alone is possible, and B & C are impossible”. The second phrase is determinism. And when people are playing chess they all assume the first phrase to be true, that one has multiple moves which one may make that would be considered stronger moves while the move considered “impossible” is not impossible because the move cannot be chosen and made but “impossible” if one wants to win the game. Kasparov and his niece illustrates this nicely. It is surprising that Pike plays and enjoys the game of chess a game predicated on his determinism being false and people actually having and making choices that are up to them and chosen by them.

    Arminibot 3000

    ReplyDelete
  8. Dear Arminibot,

    Greetings from a fellow Arminian!

    Me: But let’s say there was no reason at all for Kasparov to want to make such a move.

    Thee: I would not state it as ‘no reason at all for Kasparov to want to make such a move’, rather, I would say, perhaps the reason for Kasparov to want to make such a move is not a good one if the goal is to win the game, if the goal is something else then perhaps he has good reason to make such a move. a5 would then not be impossible but less preferable.

    I agree. I supposed it just to advance the discussion. But I thought it might be helpful if I explained why I supposed what I did. It comes down to what I think choice is and how it works.

    We have competing desires. I want to eat the steak and I want to loose weight… We make choices when our reason and desires correspond. Compatiblists like to say that we only choose our strongest desire, which to an extent throws reason under the bus. But we can control the strengthen of our desires to an extent. By thinking about loosing weight, we strengthen our desire for it and weaken our desire for the steak and vice versa. The “strongest” desire isn’t predetermined. But desire is necessary for choice, it just isn‘t sufficient for choice.

    So the two takeaways I have are:

    1) please don’t take it personally, but computers don’t desire, so they don’t choose
    2) if a person doesn’t desire something at all, he can’t choose it

    Dan

    ReplyDelete
  9. 2) if a person doesn’t desire something at all, he can’t choose it

    Dan,

    So then, we never choose God as it states in Romans 3:11:

    "there is no one who understands,
    no one who seeks God."

    ReplyDelete
  10. Dear Kevin,

    Yes that's right. Without God graciously giving us the desire to believe, we would never be able to do so. But this desire doesn't make believing automatic. We have a competing desire to live for sin and self. So we have to choose.

    But you are right that without grace we cannot believe.

    Take care,
    Dan

    ReplyDelete