Argument Mechanics - The Realm of Reason Part X (Cause)

Cause

Causal arguments attempt to support causal statements - those that reduce to the claim that A causes B.

- cause may be particular ie. this individual thing caused, is causing, or will cause something - or they may be general  ie. this type of thing causes this other type.

- causes may be affirmative or negative.

We have immediate cause, proximate cause.

Causes are not exclusionary: different interests suggest different avenues of prevention, or cure, and hence different causes.

Unlike “did” and “do” causal statements, which require more proof,  “could” causal statements are established simply by getting duplicate results in duplicate conditions.

A true “could” causal statement definitely rebuts a claim that something cannot be or have been done.

Most causation results not from a single claim of immediate and proximate causes, but from complexes of factors.

When wishing to emphasise the complexity of a problem people speak of contributory causes (no one thing is responsible).

Many a perceived “difference” has turned out to be purely psychological.

Assessing the causal arguments of others, is frequently an obligation of good citizenship.  Therefore it is important to have explicitly in mind what good causal arguments look like, and to be articulate at explaining the strengths and weaknesses of those sorts of causal arguments.

Good causal arguments are (1) congruent (they state a connection between occurrence or phenomena), (2) always contain an exclusion aspect, a ruling out.

- if we connect two occurrences we have a connection but not yet a causal connection.

correlation - a repeated, regular connection between one phenomena and another.

- a correlation connects one phenomenon with another.

- correlations may be of degree.

Correlations may be positive, or direct, or they may be inverse, or negative.

Many correlations are coincidence.  Then again many correlations result not from the action of one variable upon another but from that of yet another variable.

Some correlations, though parts of a causal chain, do not count as causal because their point in the chain is not the one at which we can exert control, or is incidental to the point in the chain where we can exert it.

post hoc ergo propter hoc = “after it, therefore because of it”, an error in logic or bad causal reasoning.

A simple causal argument can be seen as built on an “if then” premise, the causal hypothesis.

Good causal arguments are twofold comparisons, (1) “before” and “after”, but requires a control group[(i) a control is matched to the material being tested in every respect except one, namely the alleged cause, (ii) a test or trial is run, (iii) if the test material undergoes change and the control does not then the change is attributable to the difference]

retrospective and prospective studies make use of natural controls.

- retrospective studies can be valuable in defining causal issues and in leading to breakthroughs, but they are rarely sufficient to settle causal issues.

- even second-best attempts at controls are better than no attempt.

The term “control” denotes not only the standard against which a supposed causal change is measured but also the whole process of monitoring and regulating the many details, which could affect the result.

Placebo effect - “effects” tend to occur for no other reason than that subjects expect them to occur.

Test should be done blind ie. the subjects not knowing which stimulus, which batch, they are sampling.

- the test should be repeated (replicated).

Avoid a possible order effect, half the subjects should be given the test material first and half given to the control first.

“fatigue” effect - senses become dull upon successive simulation.

crossover - test and control are reversed in a latter trial.

Groups which are to be compared must be comparable.

A result is said to be causally significant when the probability of its having occurred by chance falls below a certain level.

Significance will be higher the greater the difference between control and test groups. the greater the number of trials, the greater the number of individuals and the less the inherent variability in the material.