You are viewing the historical archive of The Philosophy Forum.
For current discussions, visit the live forum.
Go to live forum

Formalization of Causation

A Son of Rosenthal July 10, 2017 at 06:20 5700 views 13 comments
Hello!
I am concerned with how to formalize some definitions of causation. In David Lewis' groundbreaking work entitled 'Causation', a definition is given as follows:
c causes e iff if it were the case that it is not the case that c occurs, then it would be the case that it is not the case that e occurs. (tense can be ignored here)

I think that so-called Lewisian analyses can be best understood as formalization of 'Ramsey-Lewis method'. It is articulated as follows:
(?x)(?y)(x causes y iff if it were the case that it is not the case that x occurs, then it would be the case that it is not the case that y occurs.)
The whole formulation can be Carnap sentence as:
If (?x)(?y)(x causes y iff if it were the case that it is not the case that x occurs, then it would be the case that it is not the case that y occurs.), then c causes e iff it were the case that it is not the case that c occurs, then it would be the case that it is not the case that e occurs.

I ask whether I am right about it here. Thanks!

Comments (13)

SophistiCat July 10, 2017 at 06:50 #85017
I am not sure what you are trying to do here. Are you saying that "c causes e iff..." is a theory and the same sentence with x and y standing for c and e is its Ramsey sentence?
A Son of Rosenthal July 10, 2017 at 07:05 #85018
Reply to SophistiCat
I am examining correct formalization of Lewis' counterfactual analysis of causation. Lewis gives a semi-formal analysis, a combination of schematic letters (e.g. c, e, ...) and English expressions (e.g. causes, ...). What I am looking for is a formal analysis of causation symbolized completely down to schema.
SophistiCat July 10, 2017 at 07:36 #85021
Reply to A Son of Rosenthal By formal analysis do you mean specifically the positivist approach of defining all theoretical terms through observable properties? I think it would be problematic to define counterfactuals this way. And that makes me think of a problem with this definition of causation:

If c is not the case, then something else is the case, right? What could be the case then? And couldn't some alternate state of affairs result in the same effect? Suppose we see a red ball strike a yellow ball and then the yellow ball rolls into a pocket. We would normally say that the red ball striking the yellow ball (c) caused the latter to roll into a pocket (e). But what would be a counterfactual to that? Suppose that a blue ball struck the yellow ball instead of the red ball, with the same consequences. (c) is not the case, but (e) occurred anyway. Does this mean that (c) was not the cause of (e) after all? It seems that on this definition of causation either there is no multiple realizability of effects or there is no causation - either option is implausible.
andrewk July 10, 2017 at 07:54 #85022
Reply to A Son of Rosenthal As Sophisticat points out, as soon as we start considering counterfactuals, we get into the devil of a mess. And counterfactuals is what's going on when we write 'if (something that we have observed to have happened) were not the case, then ....'

A less troubled approach is to define cause in terms of agreed initial conditions and scientific theories. Try reading this for an approach that does it that way.
A Son of Rosenthal July 10, 2017 at 08:02 #85025
Reply to SophistiCat
I am interested in logical forms of analyses. Or, I am interested in how to construct definitions in symbolic logic precisely.

TheMadFool July 10, 2017 at 09:06 #85033
xCy = x causes y
xRy = x is correlates with y
xNy = x is correlated coincidentally with y

1. xRy
2. ~yCx
3. ~(Ez)(zCx AND zCy)
4. ~xNy

Then

xCy <-> 1 & 2 & 3 & 4

:P
A Son of Rosenthal July 10, 2017 at 09:32 #85037
Reply to TheMadFool
1. Rxy
2. ~Cyx
3. ~(?z)(Czx & Czy)
4. ~Nxy
5. Show Cxy ? (Rxy & ~Cyx & ~(?z)(Czx & Czy) & ~Nxy)
6. [s]Show[/s] Cxy ? (Rxy & ~Cyx & ~(?z)(Czx & Czy) & ~Nxy)
7. Cxy A
8. [s]Show[/s] Rxy & ~Cyx & ~(?z)(Czx & Czy) & ~Nxy
9. Rxy & ~Cyx & ~(?z)(Czx & Czy) & ~Nxy 1, 2, 3, 4, Add
10. Show (Rxy & ~Cyx & ~(?z)(Czx & Czy) & ~Nxy) ? Cxy
11. Rxy & ~Cyx & ~(?z)(Czx & Czy) & ~Nxy A
12. Show Cxy
13. (?z)~(Czx & Czy) 3, QN
14. ~(Cxx & Cxy) 13, UI
15. ~Cxx ? ~Cxy 14, DM
INVALID
TheMadFool, your argument is not valid as shown above.

TheMadFool July 10, 2017 at 09:42 #85039
Reply to A Son of Rosenthal But I'm not making an argument. Perhaps the ''then'' misled you.

I'm defining causation.
A Son of Rosenthal July 10, 2017 at 09:51 #85041
Reply to TheMadFool
Ah, you defined causation by biconditional (or necessary and sufficient condition) ... I got it.
A Son of Rosenthal July 10, 2017 at 09:52 #85042
Reply to TheMadFool
However, your analysis includes free variables x and y, which is not allowed in first-order predicate logic. Of course, if you used them as names, no troubles here.
TheMadFool July 10, 2017 at 10:00 #85043
Reply to A Son of Rosenthal
(Ex)(Ey)(xCy <-> xRy & ~yCx & ~(Ez)(zCx & zCy) & ~xNy)

Thanks for pointing out the mistake. My logic is rusty. How's the above formulation
TheMadFool July 10, 2017 at 10:02 #85044
Reply to A Son of Rosenthal Or...

aCb <-> (aRb & ~bCa & ~(Ex)(xCa & xCb) & ~aNb)
A Son of Rosenthal July 10, 2017 at 10:24 #85048
Reply to TheMadFool
Two formulations are well-formed fomulae.