studietaak 7 (12,5%) Flashcards
scientific investigation of the causal consequences of Fracking illustrates 3 general characteristics of causal reasoning in science and in everyday situations
1) causal relationships are learned on the basis of information about the timing, location and frequency of events
2) testing causal hypotheses often involves doing something in the world, such as performing an intervention
3) causal reasoning has great practical significance: knowing about causes is how we can make things happen - and prevent things from happening - in the world
*causal reasoning is a central feature of science
Michotte
causal perception depends on spatial and temporal information (when there’s a gap between events we are much less likely to perceive them causing another)
- the mistake of reasoning from spatiotemporal succession to causation is called the “Post Hoc Ergo Propter Hoc” Fallacy (after this, therefore because of this)
Proximate Causes
are those that occurred more closely in time and place to the event that was caused
Distal Causes occurred further back in time or place from their effects
Correlation
besides spatiotemporal cues, we also tend to use information about correlation between events to discern causal relations. Correlation is a measure of the association between two variables.
Common Cause
3rd event that causes both other events
Spurious correlations
where 2 types of events happen to be correlated, but are not related in any interesting way, causally or otherwise
2 ideas about what causal relationships are, beyond the mere correlation of types of events
1) causal relationships are relationships of difference-making
2) physical process -> causation occurs when there is a continuous physical process connecting a cause to its effect, such as the transfer of energy.
Difference-Making Account of Causation
according to the difference-making account of causation, causes are those factors that make a difference to whether an effect happens or not.
the idea of difference-making can be made more specific with the help of counterfactual conditionals:
conditional - if/then
counterfactual conditionals - if it were the case that C, then it would be the cases that E
Counterfactual because the antecedent of the conditional is contrary, or counter to fact
Sufficient Causes
the causal condition is enough to bring about the presumed effect, but that effect might sometimes occur because of some other cause. If the occurrence of a cause doesn’t guarantee the occurrence of the effect, then the cause is not a sufficient cause
Necessary Cause
the causal condition must be present for the effect to occur, but the cause might sometimes occur without bringing about the effect. If the occurrence of a cause isn’t required for the occurrence of the effect, then the cause is not a necessary cause.
So sufficient causes guarantee their effects while necessary causes are required for their effects
- knowledge of sufficient causes empower us to bring about desired effects
- knowledge of necessary causes enables us to prevent some effects from happening
Causal Background
of two events comprises all the other factors that actually do, or in principle might, causally influence these two events, thereby also potentially affecting the causal relationship between the two events
Contributing Cause
a factor that increases the likelihood of an event occurring despite being neither necessary nor sufficient for the effect is called a “contributing cause”, or Partial Cause (more common than necessary or sufficient causes)
* a cause raises the probability of its effect:
Pr (E/C) > Pr (E/Not C)
Thinking about causation in terms of conditional probabilities also provides a way to define the strength of a causal relationship:
If Pr (E/C) = 1 and Pr (E/not C) = 0, then the cause is both necessary and sufficient for the effect, in any background(s) where this is true.
Judge the strength of a causal relationship:
Strength = Pr(E/C) - Pr (E/Not C)
! a necessary and sufficient cause will result in the maximum value of 1
! the strength of most causal relationships is somewhere in between the two extremes of perfect guarantee and irrelevance
Simpson’s Paradox
concerns how an aggregate statistical trend can differ from the individual trends that comprise it
- the paradox demonstrates the importance of considering the causal background
scientists have 2 methods to go beyond statistical information about correlation to uncover difference-making relationships
1) run an experiment (ideally controlled double-blind)
2) construct a causal model and rely on statistical information about variables of interest to make causal inferences
Linear correlations require variables with numerical values such as height/weight, where associations can occur between categorical values, such as “favorite color” and “favorite icecream”, or between numerical variables.
the strength of a linear correlation is measured as a correlation coefficient, which is a number between 1 and -1
* a correlation of 1 means that the two measurements form a perfect line on a scatterplot, such that when one increases, the other increases as well:
-1 = when one increases, the other one decreases by the same amount
0 = one measurement tells you nothing about the other
