Two psychologists, Al and Bert, observe a group of five (Cesar, Diane, Ed, Fay and Giulio). The five speak at different speeds (C > D > E > F > G). If A and B take up these individual differences, their judgments correspond to the extrovert and will be accurate (to the extent that the relative rate of conversation in this situation reflects the characteristic of the extrovert). It is possible to imagine a world in which A and B think that the order from the greatest extroversion to the greatest introversion is G > F > E > D > C, but what kind of world is it that pushes the perceptible to systematically see the opposite of what is true? It is less far from imagining a world (or situation) where perceptions are not correlated with reality. In a recently published article, I suggested that a “less dopey” alternative is quite true and correct. (Yes, I know, I too thought it was an impressive twist.) Excelsior keeps complete and accurate accounts. In psychometrics, the asymmetric relationship between correspondence and precision is called by paradox of the validity of reliability (Brennan, 2001). The reliability (conformity) of the measurement sets a limit of validity (accuracy). The measures in force must be reliable, but reliable measures may be valid. What is worrisome is that an increase in reliability can lead to a decrease in validity (Lord & Novick, 1968). In fact, you should usually be able to forego accurate and complete.

For example, instead of saying (forgive the rough simplification): “Acme is going to have an accurate database with the following customer information,” I would say, “Acme will maintain a database for all customers with the following information.” This obligation is inherent in the requirement that the information be correct, even if you do not use the word “accurate”. A correlation has no direction. Some scholars understand it in such a way that we can deduce the precision of concordance as well as we can deduce concordance from precision. But look at it in terms of probabilities. The conditional probability of compliance with accuracy given is p (conformity | Precision) = 1.0, while the conditional probability of accuracy to the given conformity p (accuracy | = 0.5. We now see that we are dealing with a case of reverse infertility. Because of the difference between baseline rates, the inverse inference of concordance to precision is lower than inference from one precision to another. The base rate of the chord is high (2/3), because a match can occur regardless of accuracy. The basic accuracy rate is low (1/3), as accuracy requires a match. So yes, as soon as we see the agreement, the probability of accuracy went from 1/3 to 0.5, but we arrived in a place of perfect uncertainty. .

. .