[과학철학] Cartwright (1989), Ch 1 “How to Get Causes from Probabilities” 요약 정리 (미완성)

     

[ Nancy Cartwright (1989), Nature’s Capacities and Their Measurement (Oxford: Clarendon Press), pp. 11-38. ]

 

 

  1.1. Introduction

  1.2. Determining Causal Structure

  1.3. Inus Conditions

  1.4. Causes and Probabilities in Linear Models

  1.5. Conclusion

 

  

  1.1. Introduction

p.11 #1

- How do we find out about causes when we cannot do experiments and we have no theory? Usually we collect statistics.

- But how do statistics bear on causality?

- Hume’s answer

- Salmon’s answer

p.11 #2

- The central thesis of this chapter is that in the context of causal modelling theory, probabilities can be an entirely reliable instrument for finding out about causal laws.

Given the right kind of background information about causal facts, certain probabilistic relations are both necessary and sufficient for the truth of new causal facts.

  1.2. Determining Causal Structure

p.13 #2

- Econometrics is the discipline which has paid most attention to how functional laws, probabilities, and causes fit together.

So econometrics is a good starting-place for a study of the relationship between causes and regularities.

p.14 #1

- Cartwright concentrates on econometric models rather than on path analysis for tow reasons.

- The first is that Cartwright wants to focus on the connection, not between raw data and causes, but between causes and laws.

p.14 #2

- The second is that the ideas of founders of econometrics were deeply committed both to measurement and to causes.

p.14 #3

- Tinbergen, Koopmans and Haavelmo vs. NBER and Keynes

p.16 #1

- Herbert Simon’s ‘Spurious Correlation: A Causal Interpretation’

- assumption

Causation has a deterministic underpinning: causal connections require functional laws.

The particular functional relations and linear.

All variables are time-indexed.

Causal laws are of the form Xₜ causes Yₜ₊△ₜ, where t>0.

p.17 #1

- plus the assumption

Causally related quantities are linear functions of each other generates a ‘recursive model’, 

like this:

Structural model

𝑥₁ = 𝑢₁

𝑥₁ = 𝑎₂₁𝑥₁+𝑢₂

...

𝑥ₙ = 𝑎ₙ₁𝑥₁+𝑎ₙ₂𝑥₂+...+𝑢ₙ

- Factors on the right-hand side in a structural equation are called the exogenous variables in that equation, or the independent variables, those on the left are dependent.

The us represent unknown or unobservable factors that may have an effect.

p.18 #1

- In structural equation, factors on the right are causes and those on the left are effects.

- This convention has long been followed in physics as well.

p.18 #2

- An early factor is taken to be a cause of a later, just in case its coefficient is not zero.

- This means that it is possible to infer whether an earlier factor is really a cause of a later factor or not just by looking at the probabilities.

p.23 #1

How The Laws of Physics Lie was generally perceived to be an attack on realism.

It is not realism but fundamentalism that we need to combat.

p.23 #2

  1.3. Inus Conditions

  1.4. Causes and Probabilities in Linear Models

  1.5. Conclusion

 

  

  

(2015.07.22.)