Inference Tests and GS

José Klingbeil
(French version)

Copyright © José Klingbeil. The author hereby grants permission to use this article in electronic form only, in part or in whole, to any person or institution for educational purposes, provided no charge is made for such use.

Preliminary remarks

"Observation" and "inference" constitute two verbal categories. Thus, they can only represent silent reality as an approximation and are only a map of these silent realities they represent: the objective level being a first abstracting level, our observations/descriptions already include some 'inferential' part.

In these tests, we don't mean to denigrate inferences compared with observations. In the introduction text of these tests, it is specified:

These two types of ideas are essential to our own functioning, but when we confuse inferences and observations, we have some troubles.

Some inferences are more reliable than some observations. For example, some so-called "optical" illusions are less reliable than some modern scientific theories to mention two extremes. I prefer the inferential knowledge of the lethal danger of death-cap mushrooms to the observable good taste they might have.

Given these few inevitable restrictions, it is nevertheless possible to speak about observations and inferences in a sufficiently reliable manner. In the asymmetrical relation between observations and inferences, observations refer to direct experiencing (of an object or a verbal statement), inferences being the rest.

Inference tests

These inference tests present a story, and then offer some statements on this story. The goal of the exercise is to determine if, according to a supposed truthful story, they can be considered as true, false or doubtful.

In other words, the story represents the objects, 'reality', etc. The true and false statements represent observations about the story, and the doubtful ones represent inferences about it.

Various types of errors

In this paper, we shall study a classification of three types of errors for these tests:
  1. confusing a true observation for a false one or vice versa,
  2. confusing a true or false observation for an inference (of any level),
  3. confusing an inference (of any level) for an observation, true or false.
In my experience, error 1 represents mostly a careless mistake or an error of comprehension of the story (semantic problem: one or several sentences of the story have not been understood).

Error 2 is often the result of a 'semantic' (verbal) interpretation of the tests, leading to arguing a definition subtlety. ESGS tests try to avoid as much as possible these interpretations, by using simple words and easily understandable stories.

In the following, we shall consider mostly error 3.

Nevertheless, every mistake is taken into account in the final result, since good answers obtained at random can compensate them. In my experience of 'debriefing' these tests, I consider probable that the automatically computed results are generally more favourable than those a manual correction would return.

Confusing an inference with an observation