Re: Subjective counterfactuals

hal@rain.org
Sun, 4 Apr 1999 21:23:00 -0700

Nick Bostrom, <bostrom@ndirect.co.uk>, writes, regarding my suggestion that it is meaningless to ask whether a replay is conscious:
> I see a couple of problems with this view (I'm not saying it is
> wrong).

Keep in mind that I came to this view as a result of facing the kinds of problems Eliezer describes, if we assume that replays do potentially create a new instance of consciousness. I would be curious to hear whether you have any ideas about resolving the puzzles Elizer raises.

> First, ethically: does it matter if one instantiates a
> suffering-state more times than necessary?

If replays cannot be detected by the consciousness, then it would not matter if a suffering-state is re-instantiated. This does not present a problem for this view per se, but it does call attention to the ethical aspects of the issue.

> Second, epistemologically,
> given observer self-selection (i.e. that you should consider yourself
> as if you were a randomly selected observer) then it makes a
> difference whether multiply instanciated programs should be counted
> more than once.

It has been suggested that your probability of finding yourself experiencing some event would be proportional to the number of times the experience was instantiated (among other things, perhaps). But I don't see how you could tell. If the even numbered days of your life were re-instantiated 1000 times, while the odd numbered days were only run once, would you be able to detect this?

> Also consider the slightly weird effect that two very
> slightly distinct programs give rise to two subjects of experience,
> but remove the slight discrepancy, and voila, there is no fact of the
> matter any longer as to whether there is one or two of them.

Yes, this may seem strange, but, after all, sameness is not the same as difference. A re-instantiation which is the same, or one which produces no detectable difference, cannot be detected. Only if there are noticeable differences would there be some way to distinguish them. This verges on tautology.

> (It's
> not hard to think of an example that combines this with the
> epistemological concern.)

Perhaps, on even numbered days you get re-instantiated with some perturbation. Those days you live out an alternative version of the day, but at the end of the day that instance is destroyed.

In this example, we are back to a one to one correspondence between instantiations and conscious experience. Each moment of consciousness is instantiated only once. So in that case it would seem meaningful to say that the various instantiations do produce consciousness.

Hal