Sunday, January 27, 2008

Robotic persons?

If functionalism is true, a robot could be, or could constitute, a person. Suppose, then, a robot is or constitutes a person, Robby. Assume that the robot's memory state persists when he's turned off until the next time he's turned on, for instance because everything is stored in non-volatile memory or on a hard disk.

Question: Does Robby exist when turned off?

In favor of a yes:

  1. If the robot is a person, rather than just constituting a person, then Robby exists when turned off, since robots, like vacuum cleaners and television sets, do exist when turned off.
  2. If Mary Anne Warren in her defense of abortion piece is right that persons are beings with a developed capacity for agency, then the robot plainly has such a developed capacity even when off. Moreover, some pro-life opponents of Warren will agree that a developed capacity for agency is sufficient for personhood, though not necessary since an undeveloped capacity will also do.
  3. Plainly, Robby exists when he is turned on, and when he is turned off and then on again, Robby will exist again. If it is impossible to have temporal gaps in one's existence, then Robby exists when turned off, too.
  4. What does it mean to be "turned off"? When I turn my Palm PDA "off", it has a clock that keeps on ticking, and that will wake it up next time an alarm goes off, or a user pressed a key. Is it "off"? The official terminology for a state like that is "sleep" or "standby". (The only way to really turn off a Palm device is to remove or run out its battery.) We can imagine that Robby, to conserve power, can turn everything off but a timer that will turn him on in a second.[note 1] Surely, he still exists when he's turned off but the timer is running. But the distinction between what is and what is not a part of a robot is kind of arbitrary. Suppose we take a robot that doesn't have that ability, but we use rubber bands to attach to it a contraption involving a big analog alarm clock rigged to press the robot's power button after five minutes. We could set things up so the alarm clock attached to the robot is set up so that when we press the power button, it starts running, and then five minutes later it presses the power button. It seems that we should say that the person persists when turned off when there is such an alarm clock. (There is an alternate supposition, which is that there are two robot persons there: Robby and Robby-plus-alarm. But then we'll be multiplying persons absurdly--there is Robby, Robby-minus-foot, Robby-plus-alarm, Robby-plus-dust, etc.) But it shouldn't make a difference to whether Robby exists while off if we remove the rubber bands and just hold the alarm in place. And it also shouldn't matter whether we take the alarm away for a minute, and then bring it back in time to ring. And there is little difference between that and our just being resolved to turn Robby on. But surely whether Robby exists does not depend on what our plan is--on whether we are resolved to turn him on. So, Robby exists when turned off, even if there is no alarm or resolution.

In favor of a no:

  1. It does not seem possible to draw any distinction between removing batteries and turning off. After all, turning off is disconnecting the batteries from the rest of the robot. It shouldn't matter for Robby's existence whether one is turning off the robot by removing the batteries physically from the battery holder, or disconnecting the contacts touching the batteries, or inducing a disconnection in the wires leading to these contacts. So Robby continues to exist even if his batteries are removed. But the story about the batteries can be modified further. It shouldn't make any difference vis-à-vis Robby's persistence status whether the power switch disconnects just the wires leading to the battery, or disconnects the wires elsewhere, e.g., around the CPU. We could imagine that the circuit is arranged in such a way that whenever the switch is pressed, the CPU is popped out, and this interrupts the circuit. Details of how turning off is implemented in the hardware surely should not matter, certainly not if one has the kind of functionalistic intuitions that are the main consideration in favor of thinking that there could be robot persons. But if so, we can imagine a radical case where the power switch disconnects all the major components of the robot. But again, there should be no difference between an electrical disconnect (especially on functionalist grounds) and an unplugging of a component. So, if Robby survives being turned off, he also survives all of his components being disconnected, as long as they could be put back together. It seems that the only functionalistically-acceptable constraint we can put on the taking apart is that the memory can be recovered. This leads to two absurdities. First, you can cut Robby up into small pieces and he will survive if the memory modules aren't destroyed. That doesn't seem right. (Certainly it isn't right if Robby is identical with the robot, but also doesn't seem right if Robby is constituted by the robot.) Second, if so, then we will survive as long as our memory modules aren't destroyed. But it is plausible that our brain's memory data persists a while after we're dead by any of the standard medical criteria--persists in the sense that future scientists will be able to recover it (after all, one can with sensitive instruments recover overwritten data from a hard drive). But at that time we're dead, and unless there is a soul (as I believe, but the functionalist likely denies), we don't exist when dead, so this is absurd.
  2. If we didn't have souls, we wouldn't survive while not alive. But being turned on seems analogous to being alive. (If this is right, then people in cryogenic storage are dead.) And Robby doesn't have a soul. So, Robby doesn't survive being turned off.

Conclusion: If Robby were turned off, he would exist (by the first set of considerations) and not exist (by the second set). Hence, a robot can't be or constitute a person. Thus, functionalism is false.

7 comments:

Mike Almeida said...

Alex,

Let me start with the preliminary question of why you think constitution and identity are distinct for material things like robots? I'm guessing it's a Lumpl worry. I'm actually not entirely sure of your particular reasons for denying that constitution is identity for human beings.

Alexander R Pruss said...

I think the robot is like the animal. So, if I'm not an animal, but am merely constituted by an animal, then likewise Robby is not a robot, but merely constituted by a robot.

That said, I think that I'm an animal, so that I came into existence when this organism did (i.e., at fertilization), and likewise if Robby were possible, he'd be a robot, and not merely constituted by a robot.

Martin Cooke said...

Lovely argument; my only worry is sadly akin to this terrible argument, due to Parfitt: a person's brain is split in two, and each half is given its own body, both of which are living human beings. Each of those is functionally identical to someone who derives from the original body more gradually. Are they the same person? Yes (they are each the same as someone who is identical with the original person) and no (they are two people), but they are still people so I'm wondering: Might not Robby exist in one sense and not in another, those two senses being shown to exist by your argument (if we assume that Robby is a person)?

Alexander R Pruss said...

I don't think there is more than one sense of existing.

As a dualist, my answer to Parfit's question is this: "I don't know which is happening. It depends on where the soul goes. At least three options are possible. The soul might go with the left half of the brain, or the soul might go with the right half, or the soul might leave the body entirely. The case as given by Parfit is underdescribed, because it only says what happens to the body, and not what happens to the soul." This is actually one of Swinburne's arguments for the existence of the soul: in thought experiments like this, the physical description of the thought experiment is not sufficient to determine how personal identity works. Hence, the physical description of the thought experiment is incomplete, and hence we are not merely physical.

Anonymous said...

Robby might exist as a person-of-sort-x, and yet not exist as a person-of-sort-y. That our everyday concept of a person is ambiguous (if not vague) might be indicated by Parfitt's scenario. Or are you arguing that since Robby does not have a soul, hence he is not a person, which would seem to be to beg the question?

Alexander R Pruss said...

I deny that Parfit's story shows any ambiguity in our concept of a person. It can be taken to show that the facts of physics underdetermine identity facts.

Martin Cooke said...

Then my worry was akin very sadly.