# Fallacies associated with causal and diagnostic reasoning

Firstly, let's clarify the main concepts. Consider the following example:

Jack is an introvert. He is shy and unworldly. He stays at home most nights watching TV.

The above description of Jack's character as a shy introvert could be considered causal data for predicting his behaviour i.e. that he spends his nights at home watching TV.

On the other hand, if spending the nights at home watching TV is seen as a possible cause of Jack's introverted nature, then this behaviour is diagnostic information about his introverted personality.

These different relations concern judgements of conditional probability P, defined as P(X/D) of some target event X (Jack's character), on the basis of some evidence or data D (Jack's behaviour).

Alternatively, if Jack's behaviour D is neither the cause nor the effect of his personality X, but both are perceived to be consequences of some other factor such as that Jack lives alone in the middle of nowhere, then D is referred to as indicational data for how Jack may behave in some other also lonely situation.

But if D and X do not seem to be related either by a direct or indirect causal link, then D is referred to as incidental.

Fallacies

People strive to achieve a coherent interpretation of the events that surround them. The organisation of events by schemas of cause-effect relations serves to achieve this goal.

But whereas with a normative treatment of conditional probability the data D and an event X can be equally informative, psychologically, causal data tends to have a far greater impact than other data of equal informativeness. So much so, that in the presence of data that evokes a causal schema, incidental data which does not fit that schema is given little or no consideration.

Example 1.

In which inferences from causes to consequences are made with greater confidence than inferences from consequences to causes.

Which of the following events is more probable?

That a girl has blue eyes if her mother has blue eyes.

That the mother has blue eyes, if her daughter has blue eyes

That the two events are equally probable.

Example 2

In which, when the same data has both causal and diagnostic significance, the former is generally given more weight than the latter in judgements of conditional probability.

Which of the following two probabilities is higher?

(a) P(R/H) The probability that there will be rationing of fuel for individual consumers in the UK in the next millenium, if you assume that a marked increase in the use of solar energy for home heating will occur during the last few years of this millenium.

(b) P(R/~H) The probability that there will be rationing of fuel for individual consumers in the UK in the next millenium, if you assume that no marked increase in the use of solar energy for home heating will occur during the last few years of this millenium.

Note that the event H has both causal and diagnostic significance:

Causal: A marked increase in use of solar energy should alleviate a fuel crisis. The direct causal relationship with R therefore indicates that it is less likely that there will be fuel rationing if we adopt a strategy for solar energy use. This reasoning makes (a) the more probable.

Diagnostic: The diagnostic implications of H increase R, because a marked increase in use of solar energy in the next few years implies there must be an impending energy crisis for this to be a worthwhile strategy. This reasoning makes (b) the more probable.

Example 3.

In which it is easier to assimilate a new fact within an existing causal model than to revise the model using diagnostic inference, in the light of this new fact.

Prediction and explanation represent two different types of causal inference. Models or schemas are often used to explain or predict outcomes which in turn are then used to revise or update the models. Model revision is an example of diagnostic inference.

The strength of causal reasoning and the weakness of diagnostic reasoning are evident from:

People over-predicting from uncertain models. For example predicting

academic performance of an individual from a brief personality sketch.

People constructing causal accounts for outcomes they could not predict.

People having great difficulty revising uncertain models to accommodate new data.

Consider the following description written by a clinical psychologist on the basis of projective tests:

Tom W. is a student of high intelligence, although lacking in true creativity. He has a need for order and clarity, and for neat and tidy systems in which every detail finds its appropriate place. His writing is rather dull and mechanical, occasionally enlivened by somewhat corny puns and by flashes of imagination of the sci-fi type. He has a strong drive for competence. He seems to have little feel and little sympathy for other people and does not enjoy interacting with others. Self-centred, he nonetheless has a deep moral sense.

What subject(s) would you predict to be Tom W.'s most likely field of graduate study?