Tuesday, May 6, 2014

Infinite regress explanations

Consider Thomson's toggle lamp—each time the button is pressed, the lamp toggles between on and off—but suppose it existed from eternity and every January 1 the switch has been pressed once, and only then. Why is the lamp on now? Consider the regress explanation: It's on in 2014 because it was off in 2013 and toggled on January 1, 2014. And it was off in 2013 because it was on in 2012 and toggled on January 1, 2013. And so on.

Hume will say that this is a complete explanation. But surely not. Surely the whole story does not explain why the lamp is on in even numbered years and off in odd numbered years.

Notice an interesting thing. The following are perfectly fine explanations:

  1. The lamp is on in 2014 because it was off in 2013 and toggled at the beginning of 2014.
  2. The lamp is on in 2014 because it was on in 2012 and toggled at the beginnings of 2013 and 2014.
  3. The lamp is on in 2014 because it was off in 2011 and toggled at the beginnings of 2012, 2013 and 2014.
And as we go down this list of explanations, our explanations get more and more ultimate. However, we can't take this to infinity. Each of the explanations in the list has wo conjuncts: a fact about the state of the lamp in year n, and then facts about the lamp being toggled in successive years. The facts about the lamp being toggled in successive years can be taken to infinity, but aren't enough to explain it. The following clearly isn't enough to give us an ultimate explanation of why the lamp was on in 2014:
  1. The lamp was toggled at the beginnings of ..., 2010, 2011, 2012, 2013 and 2014.
Can we take the first conjunct in explanations (1)-(3) to infinity? Well, we certainly can't in general say that the lamp was on, or that it was off, in year −∞, since even if such a year existed, dubious as that is, the lamp need not have existed then—it need only be supposed to exist in all finite-numbered years. So what can we say? Well, we could let the lamp state in year n be L(n)—0 being off and 1 being on—and then say:
  1. The limit of L(2n) is 1 as n→−∞ and the limit of L(2n+1) is 0 as n→−∞ (both limits over the integers only).
So if we think about how to complete our regressive explanation, it seems that it will need to be something like this:
  1. The lamp is on in 2014 because of (4) and (5).
Very good. But even if (4) were to be ulitimately explained (maybe there is some mechanism where each toggling is caused by the preceding, which according to Hume would give an ultimate explanation of (4)), it is clear that (5) calls out for an explanation as well, and so the regressive explanation just isn't ultimate explanation.

So infinite regresses aren't enough for ultimate explanations, pace Hume.

3 comments:

Brian Cutter said...

"Surely the whole story does not explain why the lamp is on in even numbered years and off in odd numbered years." Is there really a fact to be explained here? Isn't this "fact" just an artifact of our temporal coordinate system-- in particular, our choice of origin or t = 0 point. Another way to put this: the "alternative" possibility in which all the events are moved back one year is not really a distinct physical possibility, just as a universe in which all the matter is five feet away (in some direction) from where it actually is, or rotated by a certain angle, is not really a distinct physical possibility.

Maybe the response is to move away from coordinate descriptions. We might imagine that there is another lamp which is toggled every year in the same way, alternating between blue and red light. On all the "odd" calendar years it shines red and on all the "even" years it shines blue. Then we might ask: why is the original lamp always on during the blue years. And this fact won't be an artifact of any coordinate system.

Alexander R Pruss said...

Good objection, and I endorse your response. :-)

mjm said...

What would have been the case if nothing contingent existed ? (That is: consider a world as much like the actual world as possible, but where nothing contingent exists. To be as much like the actual world as possible, the interval during which nothing contingent existed would have to be as brief as possible. A world with a longer, possibly infinite, period of non existence of contingents is far less like the actual world than one with a brief period of non existence of contingents).

To be as similar as possible, the laws of nature (if there are such) would have to be as similar as possible to the actual laws of nature.

Which is closer, more similar to the actual world, given just "at some time t, there are no contingent beings" ? That is, which of (a) and (b) is true:
(a) In the closest possible world(s), very shortly after, something contingent would have existed.
or
(b) In the closest possible world(s), forever after, nothing contingent would have existed.
Surely (a) is closer to the actual world. But what could account for (a), contingent beings beginning to exist again ?
Only a necessary being, capable of creating contingent beings, from nothing.
But if a necessary being exists at W, it exists at all W's
So a necessary being exists (capable of creating contingent beings in at least some worlds).