The two-component model of behavior factors: The phenomenon of anticipatory engagement of a situationally dominant factor

Full Text

‘Sequential’ vs ‘parallel’ dual-process models  and precursors of action

The possibilities of studying human behavior as based solely on the rational motives are increasingly demonstrating their limitations. An alternative approach based on the theory of the so-called ‘dual process’ of decision-making and formation of intentions claims that social action is determined not only by rational, based on the conceptual thinking, but also by non-rational factors, often unconscious. In the previous article [5], we provided an overview of the main theoretical approaches in this area and proposed a two-component model of behavioral factors. We also noted that two alternative paradigms describe the logic of the interaction of conscious (explicit) and unconscious (implicit) factors, which are also often interpreted as explicit and implicit components of social attitude. The first paradigm, represented by the MODE (Motivation and Opportunity DEterminants) model, argues that the implicit group of factors is the main one. Implicit factors, organized in an a priori attitude, are automatically activated when the situation requires some action. Normally, individuals tend to shy away from spending additional resources on choosing a variant of action, and the combination of implicit factors directly determines the act of behavior. Only in some cases, the action is so significant for the subject that he is ready to use the resources of his thinking to consider available options. When a certain threshold of significance is reached, the ‘motivation’ factor is triggered. If at the same time the subject has opportunities for thinking (clear mind, knowledge and time to think), the second of the two necessary factors is activated. In this case, the rational apparatus is involved to additionally ‘refine’ the option of action, automatically projected by the initial implicit attitude, in an explicit form [18; 21; 22; 23; 32].

Theoretically, the most important consequence of these assumptions is the denial of significant discrepancies not to mention the conflict between explicit and implicit components of attitude. The authors of the MODE accept this circumstance, emphasizing that this is a direct theoretical consequence of the model, which finds empirical confirmation in the high level of correlation between implicit and explicit components identified in numerous studies [7; 8; 14; 15; 23; 32].

An alternative approach is represented by the RIM (Reflective– Impulsive Model) based on the assumption of a high degree of independence of components. Finding himself in a situation of the need to act, a person uses either impulsive (implicit, unconscious, spontaneous) or reflective (explicit, conscious, rational) models of choosing an action or a combination of actions. At the same time, the reflective model can ‘take over’ control from the impulsive one, and the impulsive one can influence the results of the reflective work. Both models have their own laws and compete for the ‘right’ to have a decisive influence on the course of action. In terms of the structural theory of attitude, in the RIM the conflict between explicit and implicit components of attitude is not only theoretically not excluded but, on the contrary, is directly predicted [36; 37]. The differences between ‘sequential’ and ‘parallel’ paradigms are schematically shown in Fig. 1 and 2.

In theory, adopting a ‘sequential’ pattern means leveling the differences between attitude components and, more importantly, denying the very possibility of conflict between them [30; 31], which entails a range of inevitable methodological consequences. Once the question of this conflict is removed, the task of studying their nature in order to search for possible sources of such a conflict disappears. The entire problem of the conditions determining such a conflict, its typical forms, ways of resolving, etc., disappears. In the instrumental perspective, the recognition of the sequential paradigm means that measurements obtained by explicit and implicit methods actually measure the same thing, and the visible differences are generated not by the special nature of measurement but by the specific method used (i.e., they are artifacts).

Fig. 1. The influence of attitude components on behavior in the ‘sequential’ paradigm

Fig. 2. The influence of attitude components on behavior in the ‘parallel’ paradigm

The design of the whole study assumed that the influence of ideologically biased statements would be measured in a ‘static’ and ‘dynamic’ mode. The theoretical research model, its operationalization, variables and hypotheses were described earlier [5], in this article, we consider only a part of them necessary for examining the presented results. In the previous article, we investigated some potential evidence for the different nature of the implicit and explicit attitude components. We were guided by the assumption that such a distinction would act as an additional and significant argument in favor of a ‘parallel’ scheme. We decided to check how the components of attitude react to the influence of stimuli which should be their factors for all conceivable reasons. The valences of explicit and implicit components of attitude towards the United Russia Party were chosen as the factors of behavior — attitude components (assumed dependent variables). The so-called ‘ideologically biased statements’ such as “The United Russia Party is a party of ordinary people” were chosen as initial factors of attitude (assumed independent variables).

In the static mode, we investigated the strength and direction of initial factors: how great is their influence on the attitude component compared to the counter influence of the component on its assumed factor. The main conclusions were unexpected: our ideologically biased statements turned out to be the initial factors for the implicit attitude component. The study revealed their strong and stable influence on this component compared to a noticeably weaker influence in the opposite direction. Concerning the explicit component, the situation was not the same: it had a decisive influence on whether the respondent agrees with the proposed statement or not. These observations were reproduced, albeit with unequal expressiveness, for all seven statements used: all the obtained values for Somers’ D passed tests of the statistical significance with a confidence level not lower than p = 0.05. All this suggested that the components of attitude cannot be something uniform, the visible differences of which are artifacts of the measurement instruments. There was no chance to find any correlation between the measurement results for our two factors of behavior/components of attitude.

The second important observation of the static tests was the following regularity: the more the symmetrical association between the initial factor and the attitude, the more the role of the implicit fraction within this association. It looked like this phenomenon contradicts the model of ‘saving mind resources’. We did not see any evidence of activation of the explicit component as the strength of association between the initial factor and the attitude increases. This would be logical to expect if we follow the theory of MODE: the greater relevance of the statements for the respondent should lead to an increase in the involvement of the explicit component. But this anticipation is not supported by our data.

The general idea of the dynamic stage

In this article, we test the reaction of the associations identified to the impact of the disturbing factor — criticality of the respondent. It was assumed that the experimental group with a high level of criticality compared with the control group with a low level of criticality would demonstrate an increase in the strength of connections between the initial factors and the explicit component and a decrease in the strength of connections between the initial factors and the implicit component. Such a forecast follows from the RIM model, which claims the independence of the nature of components, their irreducibility to each other, and allows the possibility of conflict between them. In the MODE perspective, on the contrary, one would expect a high degree of consistency of changes which should occur on a comparable scale and in the same direction. The study of this range of issues corresponded to the task of testing the hypothesis (H₀1): “there are no differences between explicit and implicit factors of behavior in terms of the nature of changes in their inherent associations with the initial factors under the influence of extraneous disturbing factors”.

The test of the hypothesis “there are no differences between explicit and implicit factors of behavior in terms of the strength and direction of associations between these factors and the group of target initial factors” ended with its decisive rejection. Explicit and implicit components of attitude differ not only in the strength of connections with initial factors but even in its direction. Ideologically-biased statements significantly affect the valence of the implicit attitude component while simultaneously being influenced by the valence of its explicit component. So, the first phase of the study led us to the results presented in Fig. 3, and it is time to answer the remaining open question (dotted arrows in Fig. 4.)

Fig. 3. Measurement result: a static view

Fig. 4. Measurement model: a dynamic view

By ‘criticality’ we mean the type of thinking, an expression which in scientific and popular literature is usually associated with ‘critical thinking’. Despite the abundance of differing definitions [9; 19; 20; 28; 29], the essential feature of critical thinking is cognitive efforts activated by an individual with a sufficient level of motivation and based on the resources of rational thinking. In the sociological perspective, the following characteristics of this phenomenon can be identified:

1. Versatility — critical thinking is not associated with any specific activities or scientific disciplines [10; 17; 20; 33; 34], which means the absence of theoretical obstacles to apply this concept to the objects under consideration (electoral attitude and the degree of independence of its implicit and explicit components).
2. Unity of method, objectivity — critical thinking is based on such techniques and methods that are the same for everyone. There can be no critical thinking ‘à la Foucault’ or, say, ‘à la Glucksmann’ [16], which provides opportunities for the correct comparison of respondents on the scale of ‘more — less’ developed criticality.
3. Objective signs that reliably testify to the critical thinking application [25; 25], which, methodologically, provides opportunities for operationalization and empirical identification of these features.

The presence of the properties listed above is a good prerequisite for using the criticality parameter to identify the reaction of the interaction mechanisms between the initial factors and the attitude components to an intervention of the ‘disturbing factor’. To simplify the model, we decided to measure criticality with a dichotomous scale: relatively high (experimental group) and relatively low (control group) levels of criticality. Based on the empirical data available, the following indicators of criticality were chosen:

1. Education — it is generally accepted that the individual gradually masters critical thinking in the course of education. In any case, one of the features of critical thinking is comparing new information with one’s knowledge [6; 10]. In addition, a relatively higher education indicates that the individual has motives for developing critical thinking. Based on this understanding, we consider education as an indirect sign of a potentially higher level of criticality.
2. Awareness of the procedure for holding the upcoming elections: in terms of criticality, having such awareness would mean better conditions for using critical thinking [10]. In addition, the knowledge of the peculiarities of the election procedure testifies to a relatively greater motivation to deliberately understand the electoral situation and make a balanced choice; i.e., it is an indirect sign of the level of critical thinking.
3. Trust in political parties: an important feature of critical thinking is conscious distrust in new information and the desire to validate it [19]. The study of the preelection situation, which provided the empirical basis for our analysis, had several variables reflecting the level of trust in various public and state institutions.

Thus, the design of the dynamic stage of the study combined two sets of associations of the statement and given component of attitude for two corresponding subsamples: low vs high criticality ones. Concerning the limitations of the available empirical data, our theoretical model was operationalized with the following set of measurable indicators:

• The explicit component of the attitude was represented by the variable formed by the answers to the correspondence of the victory of the United Russia Party in the next elections to the respondent’s interests (yes — no).
• The implicit component of the attitude was identified with the GATO test [1] and measured the valence of an unconscious attitude towards the party (positive — negative implicit attitude towards the United Russia Party).
• The set of initial factors influencing the attitude was presented by a series of ‘ideologically biased’ statements about the United Russia Party, with which respondents could agree or disagree (‘strongly agree, rather agree’ and ‘rather disagree, strongly disagree”).

Ideologically biased statements

The criticality status was set by three variables which formed three couples of experimental and control groups. The first — the results of answering the direct question about the level of education. Respondents of the experimental group had several higher educations or an academic degree (N = 302). The control group consisted of respondents with secondary and incomplete secondary education (N = 1085). The second variable was the answer to the question about the procedure for electing deputies to the State Duma. The experimental group included those who answered the question correctly (N = 1328), the control group — all the rest (N = 3922). The third variable was the results of the answer to the question of confidence in political parties. The experimental group included respondents who ‘rather trusted’ (N = 1040), and the control group — who ‘rather did not trust’ political parties (N = 2456). Finally, the set of initial variables was formed:

• AExp — valency of the explicit component of the attitude
• AImp — valency of the implicit component of the attitude
• IF1-7 — consent with an ideologically biased statement (7 variables)
• CE — education criticality status
• CA — criticality status based on the awareness of the election procedure
• CT — criticality status based on a priori trust in political parties.

The empirical data were taken from the WCIOM electoral survey conducted on the eve of the 2016 State Duma elections (N = 5248). The sample standard error is 2.25 %. The poll was conducted as part of the 4^th wave in August — September 2016 by the personal interview method, and was completed 7 days before the election day [2; 3; 4]. To test associations, we chose the Somers’ Delta indicator [4] — as an optimal measure for ordinal variables [26; 27; 35].

Key results of the study: a dynamic perspective

Like the previous hypothesis H₀1, the H₀2 hypothesis was tested by the statistical relationships between the initial factors (ideologically biased statements) and behavioral factors (implicit and explicit components of the attitude). However, if for the hypothesis H₀1 we considered associations in general for the sample, focusing on their inner nature, to test the hypothesis H₀2, we focused on the changes in the strength and direction of associations in the control and experimental group. The results for each variable are shown in Fig. 5–7. All indicators passed the test of statistical significance with a confidence level of at least 0.05. The scales represent the values of Somers’ D. The strength of the association for agreement/disagreement with statements and the valence of the explicit component is represented by the X-axis. The Yaxis reflects the same for the implicit component. Thus, each point on the graph reflects the association of the statement with both explicit and implicit components of the attitude. For instance, a hypothetical point in the upper right quadrant would indicate strong associations with both components, while a position in the upper left would reflect a strong association with the implicit component and a weak association with the explicit one, etc. The associations of each statement are represented by two points: when agreement with the statement is a factor of the component valency (‘the statement affects’) and when this valency is a factor for agreement with the statement (‘the statement is affected’).

Figure 5 shows differences in associations between statements and components of the attitude for the control and experimental groups as formed by the variable ‘education’ (CE). The experimental group of relatively high criticality consisted of respondents with several higher educations or an academic degree (AD). The relatively low criticality control group is represented by respondents with secondary education (SE). For each of two groups, every statement has two pairs of values reflected their status as independent or dependent variables. Thus, four distributions of the values are displayed in two-dimensional space. For each of distributions, trends were constructed, the functions of which were selected based on the maximum completeness of the regression by the R² criterion.

If we consider each distribution separately, for example, the group ‘statement affects (AD)’ located on the far left of the graph, we can draw at least two conclusions. First, this distribution ‘shifts’ in the counterclockwise direction, i.e., on average, its connection with the implicit component is relatively closer than with the explicit one. Second, as the strength of the symmetric association grows (values move in the ‘up and to the right’ direction), the growth rates of the tightness of the association with the implicit component decreases, and with the explicit component — grows. One can interpret remaining distributions in the same way.

The first conclusion is a significant gap between associations of statements for both the experimental and the control group: statements tend to influence the implicit component and express dependence on the explicit one (Fig. 5.1). The distributions of the control group are somewhat biased in the direction of a closer association with the explicit component, but whether such a shift is of a qualitative scale is difficult to judge based on the available data. The second conclusion is the obvious discrepancy between the types of the regression function of the distributions of the same direction of influences for the control and experimental groups. (Fig. 5.2)

Fig. 5. The nature of changes in the system of associations while moving  from the control group to the experimental group: education (CE)

Fig. 5.1. The nature of changes in the system of associations while moving from the control group to the experimental group: education (CE) — comparison of groups

Fig. 5.2. The nature of changes in the system of associations while moving  from the control group to the experimental group: education (CE) —  comparison of the direction of influence

Thus, the behavior of the measured association parameters for the control and experimental groups is fundamentally different. For the experimental (with relatively higher criticality) group, the greater the strength of the symmetric association, the more it is determined by the explicit component. This is typical for both directions — ‘the statement affects’ and ‘the statement is affected’. Graphically, this is expressed in a gradual declination of the distribution trend ‘to the right and down’, mathematically — in the fact that the best regression function is logarithmic. For the control group (with a relatively low level of criticality), the opposite situation is characteristic: the higher the strength of the symmetric association, the more it is determined by the implicit variable, which also applies to both directions of dependence. It is graphically expressed in the gradual declination of the distribution trend ‘up and to the left’, mathematically — in the fact that the best regression function is exponential.

This is a rather crucial conclusion. Is it confirmed by other measurements? The corresponding data, based on the two remaining criticality indicators, are presented in Figures 6–7. Graphical comments in the form of arrows show the differences between the directionality of the relationship for each group (discontinuous) and the type of regression function between groups (continuous). The interpretation of differences is similar to Figures 5.1–5.2.

As can be seen from the data in Figures 5–7, the patterns that form the distribution trends are stable for all three variables that play the role of a ‘disturbing factor’: the regression is either simply high (R2> 0.78) or extremely high (R2> 0.93), which allows to hope that the initial data are sufficiently reliable.

Fig. 6. The nature of changes in the system of associations while moving  frm the control group to the experimental group: awareness (CA)

Fig. 7. The nature of changes in the system of associations while moving  from the control group to the experimental group: trust (CT)

The factor of criticality, anticipatory engagement  of a situationally dominant factor, an argument  for the different nature of the behavioral factors

The phenomenon of an increase in the strength of the connection between the initial factors and the explicit component of the attitude for respondents with signs of relatively high criticality and an increase in the strength of the connection between the initial factors and the implicit component of the attitude for respondents with the relatively low level of criticality is not accidental. Thus, criticality is an essential social factor of the attitude, triggering either some mechanisms of its constituting or fundamentally different ones. Although the limits of applicability of this conclusion are not yet clear, we can safely say that there is a fairly wide area where they confidently manifest themselves. In technical terms, this conclusion is vital since the clearly manifested strength of this factor allows us to make a firm conclusion about directly opposite tendencies in the behavior of the implicit and explicit component of the attitude.

The very logic of the interaction of the components with the initial factors turns out to be multidirectional. The stronger the influence of the initial factor on the respondent, the more the explicit component is activated if it is a respondent with a relatively high level of criticality, and the more the implicit component of the attitude is activated if it is a respondent with a relatively low level of criticality. The latter phenomenon has already been noted earlier as ‘cognitive reduction’ [5]. On the example of the experimental group with a high level of criticality, we can identify an opposite phenomenon of ‘cognitive induction’. The same pattern as a whole could be called ‘the anticipatory engagement of a situationally dominant factor’, implying that for different situations different respondents have one component, then another, depending on the type of the object of attitude and on the interests and personal characteristics. The mechanism of this phenomenon can be illustrated as follows: we already know that we used ideologically biased statements as the initial factors of the attitude. These biased statements predominantly affect the implicit component of the attitude, i.e., influence the attitude to the object on an unconscious, non-rational level. However, this influence is not constant.

The group with the signs of the lowest criticality is characterized by an increase in the influence of the initial factor on the implicit component and a weakening influence on the explicit one as the strength of this initial factor increases. As it were, the average representative of this group started to ‘close off’ from the rational processing of information under the influence of the initial factor, relying more on his intuitive assessment as this influence intensifies. We describe this phenomenon as ‘cognitive reduction’.

For the group with signs of high criticality, on the contrary, an increase in the influence of the initial factor on the explicit component is typical, while this influence on the implicit component decreases. As the strength of the influence of the initial factor increases, a typical representative of this group gradually ‘turns off’ implicit mechanisms of information perception, more actively using its rational capabilities.

It seems logical to call this phenomenon ‘cognitive induction’. The corresponding data are shown in Figure 8, where the high criticality group is represented by respondents with several higher educations or an academic degree (AD), the intermediate group — by respondents with higher and incomplete higher education (HE), and the low criticality group — by respondents with secondary education (SE).

Fig. 8. A gradual change in the situationally dominant factor  from low to high criticality

Therefore, the conclusion about whether the implicit and explicit components of the attitude are independent empirical entities, or they should be regarded rather as an observation artifact [11; 12; 13], looks pretty straightforward.

In a graphical form, this conclusion is illustrated in Figure 9.

Earlier, we found that the nature of the relationship between the initial factors and the attitude component is counter-intuitive: the initial factors predominantly affect the implicit component, and they are predominantly influenced by the explicit one [5]. This conclusion allowed us to reject the hypothesis (H₀1): “there are no differences between explicit and implicit factors of behavior in terms of the strength and direction of correlations between these factors and the group of target initial factors”. The data presented in this article inevitably lead to the rejection of the second hypothesis (H₀2): “there are no differences between explicit and implicit factors of behavior in terms of the nature of changes in their inherent correlations with the initial factors under the influence of extraneous disturbing factors”. The mentioned differences within the limits of the empirical data studied are confidently manifested and are reproduced with reasonable variability in the entire set of measurements. Thus, any researcher will be forced to reject the basic hypothesis: “there are no differences in the response of explicit and implicit factors of behavior to the same initial factors both in the static and dynamic terms”. These differences are present in the most diverse forms. At the same time, there are no opportunities to explain this phenomenon otherwise than by the different nature of these factors.

Fig. 9. Measurement result: a dynamic view

About the authors

O. L. Chernozub

Email: 9166908616@mail.ru

кандидат социологических наук, председатель научно-технического совета

S. I. Tereshkin

Author for correspondence.
Email: s.tereshkin@gmail.com

председатель правления

References

Supplementary files

Supplementary Files

Action

1. Fig. 1. The influence of attitude components on behavior in the ‘sequential’ paradigm

Download (21KB)

Indexing metadata

2. Fig. 2. The influence of attitude components on behavior in the ‘parallel’ paradigm

Download (19KB)

Indexing metadata

3. Fig. 3. Measurement result: a static view

Download (40KB)

Indexing metadata

4. Fig. 4. Measurement model: a dynamic view

Download (88KB)

Indexing metadata

5. Fig. 5. The nature of changes in the system of associations while moving from the control group to the experimental group: education (CE)

Download (62KB)

Indexing metadata

6. Fig. 5.1. The nature of changes in the system of associations while moving from the control group to the experimental group: education (CE) — comparison of groups

Download (63KB)

Indexing metadata

7. Fig. 5.2. The nature of changes in the system of associations while moving from the control group to the experimental group: education (CE) — comparison of the direction of influence

Download (63KB)

Indexing metadata

8. Fig. 6. The nature of changes in the system of associations while moving frm the control group to the experimental group: awareness (CA)

Download (52KB)

Indexing metadata

9. Fig. 7. The nature of changes in the system of associations while moving from the control group to the experimental group: trust (CT)

Download (51KB)

Indexing metadata

10. Fig. 8. A gradual change in the situationally dominant factor from low to high criticality

Download (43KB)

Indexing metadata

11. Fig. 9. Measurement result: a dynamic view

Download (28KB)

Indexing metadata