Famously Hume asked “how do we know that one thing is the cause of another?’ (not in those exact words mind you) Although we can observe events and the passage of time we can never observe causation itself. There are no “causation particles” that can be examined by science. This led Hume to develop his famous problem of induction, where he showed that causation is simply a guess, not knowledge. After observing a large number of event Y followed by event X we guess that Y is the cause of X, but since we don’t know that Y really is the cause of X we can never be sure that there might not be some future case in which Y isn’t followed by X.
Now that I have presented a different account of causation let me apply that account to this problem. Ultimately my account does not rest on assumptions about what is or is not a cause, but our assumptions about how the world changes over time. Recently we have codified these assumptions into physical laws, but it is not necessary to have a formal system such as this to make claims about causation. Our assumptions about how the world changes over time generally do not include claims about causation built into them. For example the claim that objects fall towards the ground doesn’t involve the notion of causation at all. I claim that once we have these laws we then deduce causes from them (by a process of speculating as to what was necessary for an event to occur). Thus we never have to observe causation directly to have knowledge concerning it.
Unfortunately this doesn’t solve the problem of induction, it simply changes its target. Now we have to answer “how do we know that our physical laws will be true in the future?” Even though our laws may explain all of our observations perfectly we have no assurances that they will continue to apply to our future observations. Some might answer this formulation of the problem of induction with a falsificationist line of argument (a hypotheses should be considered true if it is possible to falsify it and so far nothing has). Not everyone accepts falsification as a basis for knowledge, and I will leave a defense of it for a future date, but at least we have saved our knowledge of causation from being vulnerable to the problem of induction.