On How V. L. Talanov’s Questionnaires Measure Proximity to Each of the Psychosophy Types

In principle, a Psychosophy type can be determined in two ways. One of them, which goes back to Afanasyev, the creator of this concept itself, consists in separately measuring the strength of the four Psychosophy functions – Physics, Emotion, Volition, and Logic. They are then ranked in descending order. If this order turns out to be, for example, F-E-V-L, then the type will likewise be FEVL, and so on. But in order to measure the strength of each function using a questionnaire, one needs to know the “imprint,” or, in other words, the projection of that function onto the questionnaire items. The projections of each function onto the questionnaire items can be found quite simply – first, in a training sample consisting of a large number of representatives of all Psychosophy types (where the subjects’ types are approximately known, albeit with errors – for example, self-reported by the subjects themselves), the mean response of each Psychosophy type to each questionnaire item is calculated. Then the responses of all those types in which the given Psychosophy function is by definition strong are averaged separately from those in which it is weak. The second mean is subtracted from the first – this is the projection of the function of interest onto the given questionnaire item. In this way, the projections of each function onto every single questionnaire item are calculated. Then, in order to measure the magnitude (strength) of the function in new subjects whose type is no longer known, it is sufficient to calculate the correlation between that subject’s responses and the vector of the function’s projections onto the same sequence of questionnaire items, already calculated in the training sample. This correlation (or, more precisely, the Fisher transformation of it, which makes the distribution of the correlation Gaussian) is the final measure of the strength, or expression, of this function in the given subject. Since this is simply a number, the numbers obtained for all four Psychosophy functions can be compared with one another by magnitude, arranging them into a hierarchy (and thereby also determining the subject’s Psychosophy type).

In fact, in describing this procedure we omitted a number of important and necessary corrections that make it possible to eliminate differences in the degree of noise in the functions when they are measured, and thereby to compare them more reliably at the end of the procedure. Making these corrections is very easy, and the main problem is not there. “It was smooth on paper, but they forgot about the ravines.” The point is that Psychosophy functions measured in this way turn out to be almost mutually orthogonal both across the sample of subjects and across the array of their loadings on the set of questionnaire items (see the second table in the figure). In other words, they stand in symmetrical relations to one another. According to Afanasyev, this would seem to be how it should be, but in the Psychosophy paradigm that actually exists today, this is by no means the case. What is it like, then? It is as shown in the first table in the attached image. Volition correlates positively with Physics and Logic. And it also correlates sharply negatively with Emotion. There is nothing surprising about this – in post-Afanasyev Psychosophy it is not without reason said that the first, strong Volition “attracts” the second function to itself. It is true that this delicately leaves unmentioned that it “attracts” only Physics and Logic, and does so precisely by virtue of its positive correlations with them shown in the table, whereas Emotion, on the contrary, it repels toward the end of the hierarchy, and repels it very strongly. This is likewise indicated by the strong negative correlation between Volition and Emotion.

Therefore, when diagnosing Psychosophy types by the hierarchy of four functions positioned symmetrically relative to one another, our diagnosis will very often diverge from those Psychosophy types whose properties are now more or less generally accepted in Psychosophy. That is, our diagnosis will often miss the required target (and the criterion for hitting the target is the percentage of matches between our diagnosis and the Psychosophy types self-reported by the subjects in the control group).

What, then, should be done? We need to follow the full real obliqueness of the Psychosophy functions relative to one another, which is reflected in the highly asymmetrical mutual correlations in the first table of the figure. But diagnosis of type by the hierarchy of functions is no longer suitable in this case. The best method of accurate diagnosis that takes this real obliqueness into account is to determine types directly and holistically, from the entire complex of properties of each type, rather than from the hierarchy of Psychosophy functions (which, unfortunately, smooths out the required obliqueness). And for this we also already have statistics obtained earlier from a training sample of people with known Psychosophy types. Thanks to this training sample, we know the mean responses of each Psychosophy type to each questionnaire item. Consequently, by measuring the correlation between a new subject’s responses across the entire sequence of questionnaire items and the sequence of mean responses of the “reference” representatives of each type to the very same items, we obtain 24 correlations of different magnitudes. And whichever correlation is the highest will correspond precisely to the sought Psychosophy type to which the subject is closest in terms of psychological properties.

It would seem that this time the matter is “in the bag,” that is, finally done. But again, this is only in theory, and unfortunately we must once more, for the second time, recall our sad saying: “IT WAS SMOOTH ON PAPER, BUT THEY FORGOT ABOUT THE RAVINES.” And this time there are as many as two deep ravines. First, in order to obtain accurately the projections of each questionnaire item onto each Psychosophy type (that is, to determine an accurate set of vectors of diagnostic coefficients for subsequently measuring correlations), we need to have in the training sample people with known (personally self-reported) types, with no fewer than one hundred subjects for each type. Fewer would suffice if subjects reported their Psychosophy types with high reliability. Unfortunately, people know their own Psychosophy type with approximately half the accuracy with which they know their socionic type (proven, verified). That is, there are always many errors in the training sample that add noise to the final result. This means that the training sample as a whole must consist of no fewer than 2400 people, and in reality even more – given that some Psychosophy types are poorly represented among internet users and perhaps in the population as a whole. Unfortunately, collecting such a large training sample by Psychosophy type for one and the same questionnaire is not realistic, even on the internet.

The second “ravine” is that all of the above-described training of the questionnaire in diagnostic coefficients can, with considerable effort and labor, be carried out only for one fixed diagnostic questionnaire. And what if the questionnaire changes, if its items change? What then – create a training sample for it all over again and once again recalculate all item-level diagnostic coefficients for each Psychosophy type?

But it turns out that both “ravines” can easily be cleared in a single leap, while at the same time even substantially improving the final diagnostic accuracy.

The point is that it is entirely unnecessary to search in the training sample for projections of Psychosophy types onto individual questionnaire items. It is quite sufficient (and in fact much better) to find these projections onto all independent personality factors that form the properties of Psychosophy types. It is even possible to use factors that are not fully orthogonal, but rather an oblique system of them. The only important thing is that it be complete, so that no personality factor actually used in Psychosophy is omitted from this system. For this reason, it is better for this system of personality factors to be overinclusive, even if not all of them are used particularly strongly in Psychosophy. Those that are not used will not interfere later anyway.

How does the Psychosophy classification differ from the socionic one? In some respects it is poorer (not all socionic factors enter into Psychosophy factors in full), but in some respects – richer as well. However, this latter additional “richness” concerns only one Psychosophy function, Volition. Psychosophy Volition correlates with socionic extraversion, with Te, and with Se (in precisely that descending order), and with other socionic functions and traits – to a lesser extent. But it also correlates with two factors that are entirely non-socionic. The first of them is intelligence (it does not affect sociotypes). The second, likewise ignored by socionics, is personal ambition. Ambition is unrelated to a person’s temperament; it is predominantly not an innate quality. Ambition is formed mainly by upbringing and the circumstances of the child’s life experience in early childhood. In short: the greater the number of small independent victories achieved at that “tender age,” the higher the person’s ambition will become for the whole of subsequent adult life.

Thus, in socionics neither intelligence nor ambition affects psychotypes or participates in their formation. In Psychosophy – they do. And this is only because they affect Psychosophy Volition, noticeably strengthening it. All socionic factors are fully contained in the complete socionic type profile obtained using Talanov’s questionnaire, encoded in its 16 numbers. These numbers are quite sufficient for using them to measure three of the four Psychosophy functions, with the exception of Volition. But where are we to obtain two additional indicators that reflect the subject’s intelligence and ambition sufficiently accurately and are additionally required for measuring Volition? It turns out that they too are easily measured by a questionnaire, indeed by any questionnaire, even one not specifically intended to measure them. Every Talanov questionnaire measures a person’s level of self-evaluation, reflecting the person’s tendency to overestimate or underestimate their overall success in life and resilience to adverse situations, as well as their amount of skills and abilities. This indicator is used to correct each subject’s “raw” responses because it merely introduces noise into sociotype diagnosis – for every sociotype this personal self-evaluation can be either high or low. This personal self-evaluation is, in fact, the person’s ambition that we are looking for, and it is calculated and known in every questionnaire. And an indicator of intelligence is easily found in the reliability indicator of the measured socionic profile (in the quantity displayed in the summary by the letter “S”). People with higher intelligence navigate the questionnaire items better, less often give contradictory responses to similar items, know themselves better, and are better able to compare themselves with others. Their socionic profile therefore turns out to be more contrasting and higher (all correlations of their responses with the reference sociotype vectors are higher overall). This height of the initially measured type profile is precisely the S indicator, reflecting the reliability of the results, and it also, to a substantial degree, reflects the person’s intelligence. Be that as it may, experiment shows that Psychosophy Volition correlates positively both with this S and with the person’s level of self-evaluation measured by the questionnaire, although these two indicators are not correlated with the sociotype profile.

Thus, if we add to the complete type profile its height as well (the indicator of its reliability), and also add the subject’s level of self-evaluation measured by the questionnaire and used to correct their “raw” responses, we obtain a complete set of partially oblique personality factors that takes into account all personality factors affecting the Psychosophy functions and is entirely sufficient for accurate diagnosis of any Psychosophy type. And instead of projecting the 24 Psychosophy types onto all questionnaire items, it is sufficient to project them only onto these factors. Subsequently, we then look for correlations between these projections, obtained separately earlier and therefore already known for each of the 24 Psychosophy types, and the actual values of these personality factors in a particular new subject. This is exactly what is done in Talanov’s questionnaires, and this is - even theoretically – the most accurate of all possible methods for diagnosing Psychosophy types. And today it is used only in Talanov’s questionnaires.

Only one question remains to be resolved – how did this procedure allow us to “leap” over both of the “ravines” described above at once? Recall that the first “ravine” consisted in the training-sample size being too small to determine accurately the mean response to each questionnaire item for each Psychosophy type. When we project Psychosophy types not onto questionnaire items but onto the resulting personality factors, the sample size is the same, but the accuracy of the resulting projections is already much higher. Why is this so? Because each personality factor is calculated from all questionnaire items at once, and this produces additional averaging that reduces the level of noise. Therefore, a smaller training-sample size turns out to be sufficient for accurately measuring the projections of Psychosophy types onto personality factors.

The second “ravine,” recall, consisted in the fact that when moving to a new questionnaire it was necessary to repeat the entire process of training it from the beginning and once again obtain projections of the types onto the questionnaire items because the items became different. But now, when we use projections of Psychosophy types not onto individual questionnaire items but onto the resulting personality factors, this is no longer necessary – because personality factors do not depend on the questionnaire items; they are the same for all questionnaires. And once we have found the projections of Psychosophy types onto them using just one particular questionnaire, we can subsequently use this result – immediately and without any corrections or adjustments – in any other diagnostic questionnaires as well, with different diagnostic items.

Moreover, we can continuously improve the diagnostic coefficients of the personality factors used, refining them further – even when we move to new questionnaires. This is because, although the questionnaires may be new, the personality factors in them are the same, and the statistics on them are pooled across all questionnaires even if the specific diagnostic items in them differ. That is, the method is also self-learning and self-improving.

This is how Psychosophy types are diagnosed in Talanov’s questionnaires. And the diagnosis is very accurate. From a theoretical standpoint – it is maximally accurate among all procedures conceivable for this purpose.

Intercorrelations of Psychosophy Function Strength

Table 12. Intercorrelations of function strength (on a matrix of 584 item loadings for each of the functions) in the system of respondents’ self-reported Psychosophy types

Table 13. Intercorrelations of function strength (on a matrix of 584 item loadings for each of the functions) in the system of Psychosophy types diagnosed according to the hierarchy of 4 functions