On the Social Normativity and Non-Normativity of Responses to Questionnaire Items

As we shall see below, there are two factors responsible for the normativity and non-normativity of responses. One of them is purely socionic, associated merely with certain character traits and not at all with a person’s actual “similarity-dissimilarity” to others. It accounts for 40% of the variance in the normativity-nonnormativity of responses. Owing to this factor, a person who differs greatly from others will tend to correct their responses so as to answer “like everyone else,” whereas another person, who in reality differs less from the population mean (but, on the other hand, in the required direction), will answer much more boldly and decisively, without adapting to the general opinion.

The second factor is non-socionic. It is specifically related to how objectively a person differs in their properties from the population mean. Typical Mr. and Mrs. Middle are located at the population mean (and therefore objectively answer “like everyone else”), and they too have certain characteristic properties of their own. This factor accounts for approximately 60% of the actual normativity-nonnormativity of responses.

But we shall begin by explaining how, in fact, the social normativity or non-normativity of responses can be measured using questionnaires.

----------------—

Every V. Talanov questionnaire contains several special non-socionic indicators – auxiliary or control indicators.

Among them:

1. Sigma (standard deviation, spread width) for the values of a given respondent’s raw responses. Some prefer “radical ratings” - 1 and 5, and their spread and sigma are larger. Others prefer “moderate ratings” - 2 or 4, and their sigma is smaller.

2. Sigma of the raw type profile obtained from the testing results. This is a characteristic of the height and contrast of the resulting profile. It is larger the greater the spread in the heights of the peaks in the resulting “raw” type profile (composed of correlations between the respondent’s responses and 16 vectors of diagnostic coefficients), and, accordingly, the more semantically coherent and the more contrasting the respondent’s answers to the items of all socionic scales were.

3. Social dissimulation – it shows the mean level at which the respondent rates all of their talents, abilities, and socially approved qualities (for example, responsibility and honesty). The social dissimulation scale gathers every small trace of these preferences across absolutely all questionnaire items; therefore, it depends little on specific abilities and preferences, just as it depends little on socionic profiles (although it does depend on the latter slightly). It shows a person’s general tendency to overestimate or underestimate their socially approved qualities. It should be noted that high scores are by no means always associated with a person’s deceptiveness. High scores are also obtained by simply intelligent, capable, and at the same time fairly optimistic people who justifiably rate themselves and their capabilities highly when solving a wide range of diverse tasks. Before socionic profiles are obtained, all respondents’ answers are partially corrected for their resulting social dissimulation scores – in order to partially remove this artifactual component from them, which depends more on intelligence and living conditions than on socionic profiles.

4. Correlation of the respondent’s raw responses (the ratings 1,2,3,4,5 that they give when answering any questionnaire item) with the population-average raw responses to the same questionnaire items. These mean raw-response scores, with which the respondent’s responses are compared, are obtained from statistics previously collected for all these items. The correlation is positive for absolutely all respondents, but it may differ in magnitude. For those whose ratings gravitate toward the “average general opinion,” the correlation will be high, around 0,50. For those who are not oriented toward the general opinion (or who, for one reason or another, are objectively a “white crow” in society), the correlation will be low, around 0,20 or even lower. To make this correlation easier to work with subsequently, the so-called Fisher transformation is then applied to it – it brings the distribution of correlation values in the population to a Gaussian form.

It is not by chance that we presented precisely these four indicators above (although the questionnaire actually has more additional scales of various kinds).

These four indicators, although measured independently of one another, nevertheless all turn out to be associated with a person’s tendency either to “answer like everyone else” or, conversely, to answer questionnaire items as a “white crow of society.”

At the same time, the first and second indicators listed above are negatively associated with the tendency to answer “like everyone else,” whereas the third and fourth indicators are positively associated with it.

It is no coincidence that all four of these indicators correlate with one another (see Fig.1).

==============

SEE FIG.1 in the appendix

==============

According to Fig. 1, indicator No.4 is the center of the correlation constellation of these four indicators, while indicator No.3 is the weakest indicator (in the sense of being specifically directed toward reflecting the tendency to answer either “LIKE EVERYONE ELSE” or “LIKE A WHITE CROW”).

Therefore, from the normalized values of all four indicators, we formed an integral weighted indicator for assessing the social normativity of a person’s responses, using only the three most heavily weighted indicators, according to the formula:

F= №4-0,39*№1-0,23*№2

The properties of this integral indicator were then studied on several questionnaires covering 10 thousand respondents.

It turned out, first, that it consists of two components – socionic and non-socionic, with the socionic component accounting for approximately 40% of its variance and the non-socionic component, accordingly, for 60%.

Separately in each of the questionnaires, which together covered 10 thousand respondents, we optimized a certain linear combination “G,” composed of all 15 traits of the respondents in such a way as to produce the maximum correlation with the above indicator F.

In each questionnaire, this linear combination can partly acquire its own weighting emphases for the traits (this is caused by the fact that indicator No.4 “works” mainly on questionnaire items with sharply asymmetric responses, whose mean value differs sharply in either direction from “three,” while the set of these items, in terms of their predominant socionic profile, is specific to each questionnaire). However, despite this circumstance, the correlations of the aforementioned linear combination of traits with the previously considered primary indicators Nos. 1, 2, 3, and 4 turn out to be practically identical. Their values are shown in Fig.2:

==============

See FIG.2 in the appendix

==============

Each of indicators Nos. 1, 2, 3, and 4 has its own socionic profiles, averaged not over five questionnaires, but over all more than 60 questionnaires with a total number of respondents exceeding 50 thousand.

Now it is sufficient for us to take the algebraic sum of these profiles with weights equal to the correlations of these indicators with indicator “G” from the table in Fig.2, and we will obtain the final averaged socionic profile of the factor exclusively affecting the social normativity of respondents’ answers.

Fig. 3 shows the final socionic profiles of this resulting “personality-socionic factor of social normativity-nonnormativity of questionnaire responses”:

==============

See FIG.3 in the appendix

==============

As we can see, high values of the Si, Fi, and both Democratic functions (Qi and De) contribute to the normativity of a respondent’s answers; Ne contributes to normativity only partially. At the same time, Se, Ni, both Aristocratic functions (Qe and Di), and, to some extent, Ti contribute to the social non-normativity of questionnaire responses, that is, they “draw” the respondent’s answers away from the “average public opinion.”

Why is this so?

With Fi everything is clear – by its nature, it is intended to ensure social normativity. Why does Ti reduce normativity? Because it is for truth at any cost. In a respondent’s answers, it reduces the probability that the respondent will give an answer with regard for social norms.

Both Decisive functions do the same thing – but not because they are “for truth,” but because they place a person “above others” and thereby reduce the motivation to give balanced answers that will not “break away from the collective” too much.

The most interesting question concerns the Aristocratic-Democratic functions. Why do the Aristocratic ones reduce the normativity of responses (although the Aristocratic pole is the pole of hierarchical collectivism), while the Democratic functions, on the contrary, increase normativity? (and this despite the fact that these are functions that, in all other respects, ensure a person’s individualism and independence from the pack).

The probable answer lies in the second name of the “Democracy-Aristocracy” trait, namely “egalitarianism-elitism.” Democratic functions (egalitarian ones) are directed toward the just equality of independent, free people in society. That is, according to the conceptions of the Democratic functions Qi and De, all people are equal. But in that case, individualism should indeed be closely associated with the social normativity of one’s own behavior. According to the conceptions of the Aristocratic functions, however, people are unequal and closely interdependent, and therefore each has their own role and their own place. Masters and serfs, masons and blacksmiths, etc., - all of them differ and are even obliged to differ greatly from one another precisely because of their interdependence. Therefore, there is no point in bringing all members of society to a common denominator. However, this is only our hypothesis, providing some explanation for the observed effects.

In the next post (tomorrow), we will present lists of questionnaire items most closely correlated (by the magnitude of profile covariance) with the poles of the resulting “socionic” factor of normativity-nonnormativity of responses (which, let us recall, accounts for only 40% of the variance in response normativity).

And we will see that the pole of low normativity is exactly the pole of sociopathy (antisocial personality disorder), precisely as it is described in the psychiatric literature, in all details and particulars.

This also makes another detail understandable, one that frequently arises in various kinds of psychological questionnaires (both ours and others’).

Whenever a scale is aimed at measuring sociopathy, the distribution of the value of the “sociopathy index” or of its individual components turns out to be substantially asymmetric in the population (see Fig.4).

=============

See FIG.4 in the appendix

=============

At the end where sociopathy is low, the distribution is short, as if “compressed.” At the end where sociopathy is high, however, the distribution is stretched into a long tail, making the distribution as a whole not entirely Gaussian, “skewed.” At the same time, even in the elongated tail the distribution is always continuous, without any peaks or concentrations. Some authors concluded from this that the wing of the distribution with elevated sociopathy is, after all, a certain human abnormality, and that this is supposedly why the distribution in this part stretches out, with its tail extending substantially away from the central population hump.

However, the result we obtained makes it clear that such an explanation of the “tail” is incorrect.

All people with high sociopathy scores are absolutely normal; everyone (both with low and with high sociopathy) obeys the law of the Gaussian normal distribution, while the effect of the elongated tail in the region of high sociopathy values is merely an artifact of the measurement procedure when sociopathy is measured using questionnaires. In this part of the distribution (where sociopaths are located), the value per division of the questionnaire scale is simply different; it becomes larger as the sociopathy indicator increases. This is because, in the region of low sociopathy values, the contribution of genuinely personality-specific characteristics is comparatively small, and a person’s responses are strongly influenced by unconscious corrections of those responses toward accepted social norms. But the greater the sociopathy, the smaller the correction toward social norms becomes, and the exclusive contribution of the person’s real, purely personal characteristics increasingly begins to dominate the responses. In the region of high sociopathy, no correction toward social norms remains at all; therefore, both the internal intercorrelation of responses and their association with the person’s purely personality-related factors increase, and, accordingly, the value of a division on the questionnaire scale also increases – the scale is, as it were, stretched. In other words, in the region of high sociopathy, respondents’ answers are simply less contaminated by artifacts, which is why the scale is more elongated there.

How, then, can one explain the fact that the “Judicious,” compared with the “Decisive,” give more cautious and “centered” ratings - even when the question is in no way related to assessments of “good behavior — bad behavior”?

Several explanations can be proposed.

1. A Judicious person avoids abrupt movements in everything. An abrupt movement is fraught with consequences. Therefore, the amplitude of any reactions must be damped and adjusted to the usual average level. A Decisive person has no such motives. This is the simplest explanation.

One can try to express almost the same thing in slightly different words. A Judicious person answers “radically,” with fives or ones, only on questions whose answers they are completely certain about, which reflect something important to them as a matter of principle or reflect repeatedly accumulated experience. When answering other questions that are not so significant, however, they most often experience at least some doubts. And Ne, incidentally, does not contribute to the radicalization of responses here either, because for any question it immediately offers many possible versions, forcing one to doubt and reflect. And when a Judicious person has doubts and the question does not seem fundamentally important to them, they interpret the situation in favor of cautious action. Cautious action in this case means giving a cautious rating close to the center. If not a three, then a two or a four. The Decisive, by contrast, are much more radical in this respect and are precisely “decisive” in their ratings.

I know this from myself – being Judicious, in any questionnaire my sigma score (spread) of raw ratings usually comes out below the average level. Accordingly, the sigma (contrast) of the final type profile also turns out not to be very high, although I would seem to know the meanings of the questions better than anyone. No – S always honestly comes out from 0,20 to 0,22. And I have noticed that it is the same for most LIIs. Reliability scores S above 0,26 are, as a rule, shown only by “Decisive” respondents. Why? They doubt less and do not lower their ratings toward the presumed mean even for things that are of secondary importance to them.

2. One can also try to give an entirely different explanation, with a certain degree of hypothetical fancifulness.

A Judicious person admits into their “self” the image of the human being as such; their “self” is expansive. Experience is also interpreted expansively, with an adjustment for the experience of a person “as they should be.” A Decisive person, however, has no other experience authoritative for them except the exclusively personal, egocentric one. Therefore, no responses are leveled out or smoothed.

===

CONTINUED.

In the next post, we will show lists of questions that substantively characterize the poles of the resulting socionic factor of normativity-nonnormativity of responses (without any substantive selection of questions, based solely on purely numerical characteristics). From these lists it will be immediately apparent that the pole of low response normativity is identical to dissocial personality accentuation (sociopathy).

Subsequently, in the final, third post, we will consider another personality factor – the one obtained after removing from the responses all influences of the socionic factor of “sociality-antisociality.” We will see that after this, the lion’s share still remains – 60% of the variance – for the normativity-nonnormativity of responses, although the socionic profile completely degenerates and turns into a straight horizontal line.

We will examine whether the questionnaires contain any items (even if they do not contribute to socionic profiles) that nevertheless separately characterize these true “average people,” Mr. and Mrs. Middle.

Fig. 1. Intercorrelations of the 4 indicators of normativity-nonnormativity of responses to questionnaire items (based on the results of 10855 respondents)

Fig. 2

Correlations of indicator G with the 4 control-correction scales of the questionnaire associated with the social normativity of responses practically do not depend on the choice of questionnaire. Its mean correlations across 5 different questionnaires covering 10000 respondents are shown. Indicator G itself is an individually selected, for each questionnaire, weighted linear combination of 15 socionic traits that provides, for that specific questionnaire, the maximum correlation between G and F=№4-0,39*№1-0,23*№2.

Fig. 3

Socionic profiles of the property “Integral personality factor promoting moderation and centering of all ratings”

The factor is equal to the weighted sum of the socionic profiles of the Sigma of the respondent’s raw responses, the Sigma of the peaks of their raw type profile obtained from the questionnaire, the indicator of social dissimulation of the respondent’s responses, and the indicator of the correlation of their raw responses with the population-average responses. The first two indicators are taken in the sum with weights of minus 0,12 and minus 0,23, while the last two are taken, respectively, with weights of plus 0,14 and plus 0,75.

Fig. 4