Re: Subjective counterfactuals

Nick Bostrom (
Sun, 11 Apr 1999 15:52:26 +0000

John Clark wrote:

> I think some (not all) of your [i.e. Eliezer's] questions can be answered by having
> a definition of causality that is consistent with our everyday use of
> the term and yet produces few philosophical problems:
> If event A causes event B that means two things and two things only:
> 1) If you find event A you are certain to find event B further along a
> sequence of events.
> 2) If you find event B you may or may not find event A
> in the opposite direction.

This does not seem to be a very good definition. What does "you are certain to find event B" mean? Does it mean "you are convinced that you will find event B"? Then your definition would make causal connections dependent on what you happen to think. Or does it just mean "Event B happens"? Then the sun's rising would be caused by the Big Ben striking 2 am, for whenever Big Ben strikes 2 am the sun will soon be rising (and sometimes the sun rises without Big Ben striking 2 am, e.g. when BB is broken or before it was built, so condition 2 is also satisfied).

Causation is a huge topic in contemporary analytic philosophy. Good old David Hume had a theory of causation that looks a little like what you might have been aiming for. According to Hume, event A causes event B if and only if:

(1) A is prior (in time) to B.
(2) Whenever event A occurs, B occurs.
(3) A and B are contiguous (in space and time).

(You could extend this by saying that A is an indirect cause of B iff they are connected by causal chain of events. E.g. A caused E, E caused F, F caused G, and G caused B.)

There are several well-known problems with Hume's theory which modern theories try to overcome.

Nick Bostrom Department of Philosophy, Logic and Scientific Method London School of Economics