Problems for Hempel and Salmon

Problems for DN and ISN
Account of Laws
On DN laws are
(a) exception-less generalizations
(b) counterfactual supporting
With (a): most laws are not (a); e.g. Mendel’s law of segregation (M) has a number of exceptions such as meiotic drive (39, Woodward).
With (b):even Hempel’s example that all the rocks in the box contain iron supports the following counterfactual: if I were to pull out of the box a rock, then it would contain iron.  So, we need a much more fine-grained account of the type of counterfactual relevant.
Laws in General: DN is committed to the relevance of laws and yet we do not have a satisfactory account of laws.  
Some explanations that don’t cite laws seem perfectly adequate: the knee bumping the desk caused the water to spill
General Problems for DN: 
First: DN fails to respect the asymmetrical nature of explanation.  The height of the flagpole, the angle with the sun and so forth explain the length of the shadow.  But from the length of the shadow, the angle of the sun and so forth one can deduce the height of the flagpole.  According to DN, the shadow (with the other stuff) explains the height.  This is false.   (Bromberger, 1966)
                        In general, if x explains y, then y does not explain x.  There are no circular explanations.
            Second: All males who take birth control pills regularly fail to get pregnant.
                        JJ is a male who has been taking birth control pills regularly.
                        JJ fails to get pregnant.
This conforms to the DN model.  We have a law, a particular fact and an outcome or some event that needs explaining.  The idea that one can explain an outcome solely on the basis of a law and circumstance looks to be in trouble. 
For INS Model
INS replaces the exception-less account of laws with probabilistic laws. Perhaps all laws are probabilistic. If so, then INS, if it is true, is the correct account of all SE.
             But, we still do not have an account of what laws of nature are. 
But, INS requires that the laws (probabilistic ones now) plus the circumstances imply a high probability of the occurrence of the event to be explained, and this seems false to many.
We can explain Tim’s leukemia by citing the law that states that persons who are 2km from the blast have a 20% chance of getting leukemia together with the fact that Tim was 2km from the blast. If this is right, then INS is false. 
Problems for Salmon
The probability of Tim getting leukemia given his distance from the blast and the occurrence of the blast is greater than the probability of Tim getting leukemia given no blast or given blast but great distance.  So, the distance and the blast make a difference here.  
P(L/D&B) > P(L/~B)
L=Leukemia
D=Distance
B=Blast
The above says: the probability of getting L given D&B is greater than the probability of getting L given that there is no blast.
Thus, distance and blast are statistically relevant.  Salmon thinks this is enough to show causal relevance.  So, we have a genuine scientific explanation of Tim’s getting leukemia.  Or do we?
First, note that Steve does not get leukemia even though Steve was also 2km from the blast. How do we explain this? In the EXACT SAME WAY that we explain Tim’s getting leukemia. So, the same explanation tells us why Tim got L and Steve did not. That’s weird. Salmon's type of explanation tells us why Tim got L rather than remaining L free (the probability of him getting L given blast and distance is greater than the probability of him getting L given no blast) and it tells us why Steve did not get L rather than getting it (the probability of him get L given blast and distance is less than the probability of him getting L given blast and distance--P(L/D&B < P(~L/D&B). Notice that the probabilities are true of both Tim and Steve. It seems like we do not really have an explantion for Tim's getting L in such a case. 
     Note that this is parllel to the driving over the ice example I gave earlier.  
Second, it seems that there are (or can be) causal processes that are instantaneous—quantum events.  This would imply that Salmon’s account of a causal process won’t work since it will fail to distinguish between causal and pseudo-causal processes. I am not sure that this is a big worry, but it might be. The devil is in the details!
Third, I am pretty sure that hardly anyone thinks that causal relevance can be explained solely by statistical relevance (I think Phillip Dowe and John Woodward have shown this to most people’s satisfaction but I don’t presently recall the exact sources—Dowe is super fun to read on relativity and quantum stuff—if you are interested I can send out some stuff he wrote on God’s relation to quantum events).
Here’s why:it is possible to have statistical relevance without causal relevance. The basic idea is really just a twist on the idea that correlation is not causation (by the way there are some who doubt this but who still think that Salmon’s analysis fails).  
Question: Can you come up with such a case? That is, a case where we have statistical relevance without causal relevance? I'll post one later :-). 




Comments

  1. M. E. G.February 18, 2019 at 3:01 PM

    A case with statistical relevance but no causal relevance?? Hmm... Could it be along the lines of statistically we can measure the amount of times we are sad compared to when we cry but cannot figure out the causal relevance of tears?

    ReplyDelete
  2. DAFebruary 19, 2019 at 7:18 AM

    Maybe, but our ability to figure out the causal relevance is not the main issue here. What we need is a case where something, x, is statistically relevant to something, y, and yet x is not causally relevant to y. Note also, the case does not have to be actually true to show that stistical relevance and causal relevance come apart. Here's an abstract representation of what we are looking for:

    If x is statistically relevant to y, then P(y) < P(y/x). That is, the probability of y occuring is less than the probability of y occuring given that x occurs. Put slightly differently, P(y/~x) < P(y/x). That is, the probability of y occurring given that x does not occur is less than the probability of y occurring given that x does occur.

    Example: Is applying heat statistically relevant to water boiling? Yes! water boiling in the absence of heat application is less likely than water boiling in the presence of heat application.

    So, what we need is this: a case where we have P(y) < P(y/x) but where x is not a cause of y.

    Does that help?

    I think your crying and sadness might be made to work. Give it a try!

    ReplyDelete
    Replies
    1. DAFebruary 23, 2019 at 12:20 PM

      Quick, minor correction (this is does not substantially change anything I said above): If x is statistically relevant to y, then P(y) # P(y/x), where # stands for 'is not identical to'

      Delete

Post a Comment

Popular posts from this blog

Lakatos and Thagard: Demarcation Problem

Lakatos-Falsification and Methodology Naïve Falsificationism (NF): Rules of Acceptance:  if T is falsifiable, the T is scientific. Rules of Falsification:  If some observation statement conflicts with T, then T is falsified. Sophisticated Falsificationsim (SF):  Rules of Acceptance:  T is scientific only if T has more empirical evidence than some other T* Rule of falsification:  T is falsified iff some other theory T* has greater predictive and explanatory power than T and is confirmed to some degree. Theoretically Progressive:  T is TP if T purports to take in more empirical data than some other theory T*. Empirically Progressive:  T is EP if T is TP and some of the excess empirical data is confirmed to some degree. Progressive Problemshift:  both TP and EP If a Problemshift is TP, then it is scientific, if not TP, then it is pseudoscientific So, some theory T is  pseudoscientific  if T is an alter...
Read more

Diverse Viewpoints: Should we be open?

There are lots of different factors in not only theory acceptance but also in accepting and maintaining various beliefs. One is acquiring relevant information about the topic. Another is acquiring and assessing different interpretation of that information. Another is comparing the theory or belief with other theories and other beliefs to see if they fit together well. In some cases, embracing a new idea or theory requires changes in one’s existing beliefs, and sometimes those changes seem too drastic. If embracing theory T means I have to change my belief that 2+2=4, then I ain’t embracing T. That seems perfectly rational. What theories should we really seriously consider? By that I mean, give them serious thought, and be open to changing our minds. Are there ideas, views, theories, that we should not give any serious attention to, even if they have defenders?  We mentioned the Alien-laser theory of drunk driving, but that is not defended by anyone as far as I know.  ...
Read more

Defeaters

Undercutting : where  E  is evidence for  H , an undercutting defeater is evidence which undermines the evidential connection between E and  H .  Rebutting : is evidence which prevents  E  from justifying belief in  H  by supporting not- H  in a more direct way.  In general: a. Moving from believing that p to believing that not p requires a rebutting defeater b. Moving from believing that p to not believing that p requires an undercutting defeater Note the difference between the scope of the 'not' in a. and the scope of the 'not' in b. In a. one believes something, namely, not p. In b. one does not believe something. The 'not' in b. is negating the believing, whereas the 'not' in a. is negating the thing that is believed.  Ok, given all of that, you should be able to explain the relevance of the following cases that Plantinga brings up in Chapter 6. Are they examples of undercutting or rebutting defe...
Read more