Problem of Adequancy of Sociological Investigation Results

Sociological and marketing research study the attitudes of various group informants towards some objects, positions, opinions or statements (A and B) that present an interest to the subject arranging the investigation. When forming a sampling, it is appropriate to split the informants by different mass media audiences, because it is easier to further influence each group separately by means of mass-media.

Each group would interpret given objects according to the attitudes of its shared mentality. Also, each group can give its interpretation variants determined by individual, territorial, linguistic, cultural, educational and other differences. They give additional data dispersion for each social group of informants. It seems impossible to minimize estimates dispersion and, at the same time, increase their validity at the account of individual factors, because it requires additional psychological testing for each group with the help of an indefinite number and scope of tests. Besides, one and the same interpretation variant can be grounded by a different grouping of individual and social factors.

EXAMPLE

Lets assume there is a certain marketing research based on two social groups industrial workers and office personnel. To make it simple, lets view only two factors that cause the different estimates - social, and generic. In our case we are supposed to obtain interpretations for a definite object of research: A-1, B-1, and A-2, B-2 (see Fig.1)

Fig. 1

The interpretation variants A and B expected by the investigator are obviously extended in each mentality segment by their specific semantic elements (to receive A-1, B-1 and A-2, B-2). The investigator may not suspect the existence of these variants; neither can they appear in traditional sociological (marketing) research, while all the factors defining the mentality are previously obscure.

Fig. 2

This means that even though the informants may give identical responses to the questionnaire points, those eventually may have a different sense.

Hence, statistic tables traditionally used in sociological and marketing research, don't yield an adequate description of any consumer choice (mentality segment).

For instance, while determining a percentage of informants who consider a certain product high quality, it is seldom found out, what "high quality" means concerning this class of products within a given mentality. Even if such analysis ever takes place, as a rule, it is inadequate is inadequate.

A formal estimate of an average opinion can lead to paradoxical results. Taking example A1, the informants can see with equal probability both 6 (interpretation 1) or 7 (interpretation 2) cubes, therefore, the average estimate (6.5 cubes) represents no actual opinion of the surveyed informants.

The method of Semantic Analysis proposes an adequate solution to this problem, while the informant's interpretations are originally laid out in a multi-dimensional semantic space.

Such an approach is especially important in the estimation of desirability (rating) of objects under study. They evidently depend on their specific interpretations and probabilities in different mentality segments. The rating dispersion in each case will be determined by the corresponding probability ratings of obscure interpretations. Therefore, the traditional approach will require large samples of informants, which promises no valid result.

Besides that, there can exist polar reactions of acceptance and rejection, the interpretations being identical. This brings additional dispersion to the survey results. Increased dispersion may account for the sufficient polarization of society concerning obscure points of the surveyed object.

Thus, increase in result validity is reached at the expense of lowering the estimate dispersion, which, in its turn, is arrived at by adequately segmenting the mentality. The latter is calculated by selecting similar meanings in the informants' responses, and never by selecting formal, investigator-preferred properties or mechanical increase in informant number.

Semantic Analysis consecutively identifies the most probably correlating opinions within a given mentality (starting from maximum probability), and therefore, it is not critical to the number of subjects surveyed (sampling). In fact, a researcher can restrict the sample either by a number of most frequent positions (opinions), or the limiting probability of a position, or the combined forecasting error. As a rule, satisfactory accuracy is reached when taking only 2-3 first positions (states), for which 30-60 informants are enough for a studied mentality segment, while their contribution to the rating drops almost exponentially.

It should be noted that position probability is never determined by arbitrary factors, but it has its own psychological basis and experimentally testable grounds.

As far as the sociological surveys don't disclose attitudinal mechanisms (algorithms) for decision-making within a definite mentality, they have to conduct a continuous monitoring of various situations, being incapable of scientifically grounded prediction. Researchers employ regressive probability approximations out of despair, because the latter are purely empirical and devoid of theoretical sociological or psychological basis. This can not be called sociology, because it's pure statistics.

Back