An experiment in sociology. Social experiment in sociology

An experiment is an experimental study of the effect of a single factor (or several factors) on a variable of interest to the researcher. An experimental study is built in accordance with the rules of inductive inference about the presence of a causal relationship between events, firstly, demonstrating the regular nature of the occurrence of an event after the event-impact preceding in time, and, secondly, excluding alternative methods of experimental isolation and control. explanations of appearance with the help of extraneous influences and competing causal hypotheses. Accordingly, experimental data represent the best approximation to the model of statistical inference about the presence of a causal relationship between exposure and, in more familiar terms, between the independent and dependent variables.

Definition and types of experiment.

Basic Principles of Experimentation in the Social Sciences

The experimental method came to the social sciences from the natural sciences, where it became the main method of experimental verification from about the 17th century. scientific theories... The most popular type of experiment in the exact sciences has been and remains a laboratory experiment, which has also become widespread in the sciences of human behavior.

A laboratory, or true, experiment is aimed at testing a theoretical hypothesis and is carried out under conditions of maximum control over the level of influence of the independent variable and purification (isolation) of this effect from extraneous influences exerted by external, i.e., irrelevant from the point of view of the hypothesis being tested, by variables. Experimental control and isolation allow one to reject other possible explanations for the observed effect - competitive hypotheses. An important condition for the validity, validity of the results obtained in laboratory experiment, is the possibility of a sufficiently reliable measurement of the dependent variable. In this case, with an infinite number of tests, the results of inevitable random perturbations in the dependent variable of each other and the researcher will receive an accurate estimate of the effect of interest.

In practice, the described requirements for a true experiment can be fully implemented only in an infinite ideal experiment, during which the external, so-called exogenous variables remain unchanged, and only the independent variable changes, which ensures the full validity of the conclusions about the studied relationship between the independent and dependent variables. An ideal experiment is a standard against which real experiments can be evaluated and compared, but literal fulfillment of all its requirements is usually impossible or even meaningless from the point of view of a specific scientific task facing a researcher.

In the social sciences, it is customary to distinguish between a field experiment conducted in a laboratory and a laboratory experiment. natural conditions and in most cases, its purpose is not so much to test a scientific hypothesis about a causal relationship between variables, but to evaluate the effectiveness of various programs or methods of influence.

For many applied sociological research focused on the development and evaluation of social programs, social experimentation is as typical as laboratory experiments for social psychology or small group sociology. Social experimentation provides answers to a wide variety of questions, predominantly in the realm of practical policy and administration, such as how cancellation affects death penalty on crime rates, whether museum attendance grows with decreasing entrance ticket prices, whether in all cases an increase in remuneration leads to an increase in labor productivity, etc.

For example, in a study of the impact of a children's TV series on the cultural and intellectual development of American preschoolers, the field experiment involved children and parents living in cities (Boston, Durham, Phoenix), as well as in rural California and Philadelphia. During the experiment, children and their parents were encouraged to watch the series (independent variable), recording changes in the cognitive development of preschoolers using achievement tests and tests of general development (dependent variables). A two-year field experiment demonstrated a remarkable educational effect associated with watching the series, especially evident in a group of disadvantaged children.

Field experiment is the leading method for practice-oriented evaluative research. R. Gottsdanker proposed to distinguish between two types of field experiments - experiments that duplicate the real world (i.e., experiments already described), and experiments that improve the real world. Experiments that improve the real world primarily improve the validity and reliability of data. Thus, experimental data on the effectiveness of a new method of teaching driving a car will be influenced by many hard-to-control differences in visibility, terrain, road conditions and vehicles used, while simulator data will be less susceptible to such biases. In addition, the reliability of the experiment under conditions will also increase due to the ability to control the frequency on the simulator screen, observe a clear time regime that excludes operator fatigue, etc.

In order to understand what guides sociologists when choosing one of the listed types of experiment and a specific plan for constructing experimental research, we should first consider the key concepts and principles that underlie the modern approach to social experimentation.

Summarizing the above, we can say that in a laboratory experiment, the validity of the conclusions of the experimental study, i.e. their validity and reliability are ensured thanks to three principles of experiment planning: 1) control over the level of the independent variable, 2) isolation of the main effect (i.e., the actual effect of the independent variable on the dependent variable) from the influence of extraneous, confusing factors, and 3) multiple reproduction of the results obtained, which makes it possible to neutralize random changes in the result of individual tests associated with unsystematic background fluctuations, random errors, fatigue, etc. In this case, the first two principles of planning a laboratory experiment make it possible to ensure validity as the correspondence of the experiment to its purpose, the measurement of exactly that effect, which was supposed to be measured. An ideal, that is, a perfectly valid experiment, fixes only the relationship between the variables that the experimenter plans to study, and any other sources of systematic variation in the results. The validity of the experiment, therefore, determines the reliability of the conclusions about the presence or absence of the alleged causal relationship and about the confirmation or non-confirmation of the theoretical hypothesis tested in the experiment (Fig. 1). The third principle ensures the reliability of the results - protection against accidental errors, which, as mentioned above, is a necessary condition for validity.

However, most of the experiments in the social sciences (as, indeed, in a number of engineering disciplines or agrobiology) take place in conditions where the listed principles cannot be fully implemented. The limitations arising here are of a technical and sometimes more fundamental nature. If, for example, in a sociological experiment, race or social origin serve as independent variables, then we cannot even technically - not to mention moral considerations - exercise complete control over these variables, i.e., completely arbitrarily determine their value for each individual case ( subject). And even if this difficulty could be somehow overcome during the test, it would be difficult for us to isolate the main effect of interest to us from the influence of concomitant factors, with an independent variable, related to race or origin of specific social experience, corresponding social skills. etc. (An agrobiologist is faced with similar difficulties when assessing the yield of a new wheat variety and trying to separate the main factor from others that also affect productivity: differences in the illumination of experimental plots, in the composition of the soil, etc.).

The famous English statistician R. Fisher was the first to substantiate the possibility of using a slightly different approach to planning field experiments, laboratory experiments with incomplete control, as well as quasi-experiments (the latter will be discussed later). This approach is based on the purposeful use of the laws of chance and the theory of probability. It requires the introduction of the principle of randomization into experimental design.

Randomization is a strategy for randomly assigning subjects to different experimental conditions (modes) and experimental groups.

We will use as an illustration the same simple example of psychophysical experience that Fischer himself uses when discussing the role of the principle of randomization in experimental design.

Imagine a lady was trying to guess which four of eight cups of milk tea were first poured with milk, and which ones were filled with tea. The experimental hypothesis is, therefore, that the subject possesses the described discrimination ability. Using the simplest combinatorial formula, the total number of ways to select four cups out of eight is equal to the number of placements A:

In other words, if the subject did not have the ability to distinguish cups from and resorted to random guessing, then the probability of correctly identifying 4 cups would be 1/70 (out of 70 existing different ways to choose 4 cups out of 8, only one is correct). More precisely, with a sufficiently long series of tests, the frequency of correct answers would approach 1/70. If the frequency of correct answers significantly, from the point of view of the chosen statistical criterion, exceeds the frequency of random guessing, then we have the right to conclude that the special ability of interest to us really exists.

However, the described test scheme clearly does not meet the requirements for the level of control over variables and isolation of the main effect in a laboratory experiment. The potential threats to the validity of our statistical conclusions are fairly obvious. First, we do not have any means of directly measuring the dependent variable, the ability to taste taste. Correct guesses are only indirect indicators of this ability and may reflect the influence of variables, even exceeding the main effect. If, for example, sugar has been added to all the cups in which the milk was first poured, it is clear that they will all be unmistakably identified. Statistical conclusions about the significance will thus turn out to be invalid, i.e., not relevant to our experimental hypothesis (although they would confirm the competitive hypothesis about the ability to recognize sweet taste). A systematic extraneous influence that threatens the validity of a statistical inference can also be of a different nature: the cups can differ from each other in thickness, temperature, and color. In order to eliminate the listed threats to validity and level the arising systematic biases, the experimenter must use the principle of randomization, i.e., ensure a random order of presentation of stimuli-cups and their assessment, determined, for example, using a table of random numbers or drawing lots. Ideally, both the selection of subjects for each series of experiments and the distribution of the levels of the independent variable between the dishes (or) should be based on the laws of chance.

When planning a specific experimental study, the described principles are embodied in the development of a plan, or scheme, of an experiment that determines the order of presentation to the subjects (or their groups) of different levels (conditions) of the independent variable to adequately test the experimental hypothesis.

Basic experimental designs with control group and randomization

In sociology, psychology and other behavioral sciences, a special role is played by the use of the principle of randomization in the distribution of subjects into groups. In an experiment with incomplete control or in a field experiment taking place in natural conditions (i.e., in the conditions of a school class, industrial organization, urban area, etc.), it is often necessary to prove not only the existence of the expected effect as a result of some influence of X, but and lack of the same effect in cases where there was no exposure. For example, a researcher studying the effect of watching anti-war films on changes in student attitudes randomly selects an experimental group from a population of students to be shown an anti-war film and a control group to which he will show a neutral film that has nothing to do with the studied attitudes. The design for this simplest randomized experiment with pre-test, post-test and control group (RT1-2C) would look like this:

where R is the procedure of randomization (random distribution into groups), O1.2 are the levels of attitudes in the experimental group before and after watching the movie X, O3.4 are the levels of attitudes in the control group that did not watch the film. The use of a control group eliminates some of the critical threats to the validity of the experiment. First, if the researcher refused to use the control group and limited himself to testing, that is, measuring viewing attitudes, then the observed changes in the level of attitudes could be attributed to the influence on the subjects of the very fact of participation in the experiment. The subjects may have realized that they were selected for important research and sought in some way to match their role and justify some hypothetical experimenter. The described threat to validity is widely known and there are several designations for it. In psychological testing and experimental psychology, it is called or. Sometimes they use a term that originated in medicine, where, in clinical trials of new pharmacological agents and methods of treatment, it is often observed, that is, a noticeable improvement in the status of many participants in the control group, in which, instead of a real effect, indifferent means and neutral medical manipulations were used. In sociology, the most popular designation for systematic bias arising from the reaction of subjects to the experimental situation is this.

In the so-called Hawthorne experiments (after the name of an industrial plant in Chicago), organizational and socio-psychological factors influencing labor productivity were investigated. The researchers found that the effect of productivity growth in teams persisted even in the absence of actual experimental intervention. The presumptive reason for this phenomenon was the growth of group self-awareness among the participants in the experiment.

This is a validity threat associated with the characteristics of the experimental group.

Another threat to the validity of conclusions that the described experimental design allows to overcome is also associated with the characteristics of the groups, namely, with the selection process for participation in the experiment. If we had abandoned preliminary testing and random distribution of subjects into experimental and control groups, we would have made a methodological error, which is very typical for any experiments with volunteers. Very often, researchers select subjects to participate in a special training program or in an innovative organizational project based on their expressed desire, and then compare the results shown by the participants in the assessed program with the results of some other available group (or even a random sample from the corresponding population). ) that did not participate in this kind of program. However, such a comparison is incorrect: the desire to participate in an experiment itself often indicates an initially higher motivation, awareness, or intelligence. These factors, by themselves, or interacting with the main independent variable X, can explain the significant differences in the results shown by the experimental group. In the example just described, comparing a group of volunteers who wished to watch an anti-war film with other students may overestimate the impact of viewing if the volunteers are initially more interested in political issues. Only a random distribution of volunteers in the control and experimental groups, while observing such a distribution for all subjects (participants in both groups should consider that they are exposed to some experimental influence), makes it possible to judge the role of the independent variable X in the occurrence of intergroup differences.

The above threats to validity are mainly associated with the characteristics of the groups, manifested at the stage of selection or during the experiment. However, randomization makes it possible to cope with some validity threats posed by variables of the external environment, background. The background includes, in particular, as well as factors of natural development -. The subjects grow up, learn, get tired, improve their results with repeated tests, etc., which can affect their results. However, if such extraneous influences do not selectively affect only the members of the experimental group, they will only contribute to random error, and not to systematic bias. In other words, they will be equally likely to be distributed between randomly selected participants in the control and experimental groups. The RT1-2C type design allows for a random and equiprobable nature of external, background influences on the control and experimental groups. Moreover, it allows the magnitude of the background influence and to assess the net main effect: if the external influence did take place, it equally affected the indicators of both the experimental and control groups; therefore, the difference between the mean values of the first and second measurements of the level of the dependent variable in the control group (O4? O3) must be subtracted from the same difference in the values of the dependent variable recorded in the experimental group (O2? O1), i.e.:

NS? = (O2? O1)? (O4? O3),

at the same time, preliminary and post-experimental testing in the experimental and control groups should be carried out practically simultaneously.

A systematic threat to validity associated with background factors can still arise with the use of randomization and control groups. This occurs when background factors interact with the independent variable (or some of its levels). The nature of this interaction is easy to understand through the example of a study that examines the effect of anxiety arising in a situation of uncertainty on the success of a solution. difficult tasks... In experiments of this kind, to create a situation of uncertainty and increase reactive anxiety (independent variables), they often use vague instructions, indirect negative assessments of the subject's actions, expressed by the persons conducting the experiment (such as), as well as preliminary series, where the subject has to solve obviously unsolvable problems. Of course, only members of the experimental group are exposed to all these influences (X). If the subjects are students who, due to a random coincidence of circumstances, have to take exams in a week, or employees of a firm's division awaiting an early re-certification, then these factors will interact with the independent variables, not only adding up with them, but also enhancing their effect. Negative reinforcement received during the experiment will be perceived much more sharply on the eve of the examination session or re-certification, and the background anxiety and uncertainty associated with these events will interact with anxiety and uncertainty created intentionally.

In this case, it is impossible to assess the net effect of interaction by comparing with the results of the control group in this case, since the interaction of background displacements with the main effect occurs only where there is an experimental effect X. In the control group, an approaching unpleasant event can also affect the results of the final measurement or even both measurements, but it will not interact with the effect of the independent variable (such an interaction can be described in more precise terms, but this requires the introduction of some statistical concepts). To estimate the magnitude of the interaction of background mixing variables with the main effect, more complex experimental designs are needed. Some of them will be discussed below.

Sometimes it makes sense to use a simplified version of the described plan with randomization and control group, namely the plan with randomization without preliminary testing (RTC), which schematically looks like this:

The attractiveness of this plan lies primarily in its cost-effectiveness. If a true probabilistic procedure was used in the distribution of subjects between groups and levels of exposure, then conducting preliminary testing - as in the KT1-2S plan - only increases the cost of the experiment without significantly affecting the quality of the data obtained. In fact, correctly performed randomization (R) is the best guarantee that there is no initial bias between groups (i.e., equality of the initial group mean values of the dependent variable O in the control and experimental groups). In addition, in large-scale sociological research, as well as in applied research in the field of pedagogy and social work, conducting both preliminary and repeated measurements of the dependent variable (political attitudes, criminal behavior, etc.) is often unfeasible or leads to a reaction to itself. testing procedure. If, for example, we study the impact of participation in a pre-election group discussion on subsequent support for political parties, then a preliminary measurement of political attitudes can affect the activity of respondents during the discussion itself and their subsequent attitudes.

It can be concluded that the RTC plan is always preferable when it is undesirable to re-present tests, questions, and measurement procedures that are very specific in content and similar in form. In other words, this plan deserves widespread use in sociological and evaluative research. In addition, it is the most acceptable means of conducting so-called methodological experiments, in which the effectiveness of various types of surveys, methods of filling out and formats of questionnaires, etc., is assessed (an example of such a methodological experiment is the study of Bradburn and Sudman, described in the chapter, in which the effectiveness of the method compared with the effectiveness of the use of administrative statistics data). The RTC plan was used, in particular, in one of the practically oriented social experiments to improve the system of law enforcement and penitentiary institutions, conducted in the United States in the 1960s and 70s.

In 1961, the US Federal Bureau of Penitentiaries began a three-year experiment to examine the effects of supportive counseling and special rehabilitation programs on the behavior of adolescent boys in prisons. The place of the experiment was a special colony school for juvenile offenders.

In the course of randomization, boys were randomly assigned between the teams where the experiment was conducted and the teams that served as control (each team lived in a separate building). In the experimental groups, the number of educators was increased, special individual and group consultations were conducted, and a system of rewards for good behavior was used. The control groups used the usual methods of education and training, as well as traditional punishments for violation of the internal order. The results of the experiment showed that boys from the experimental groups left the special school earlier, did better in their studies, and behaved more adaptively. No significant differences were found in the rates of recidivism for adolescents from the experimental and control groups released from the colony school, however, the children from the experimental groups differed significantly from the children from the control groups in terms of the severity of newly committed crimes (the former, in the case of a recidivism, committed less serious crimes). Based on the results of the experiment, it was decided to introduce an experimental correction program in all departments of the school.

Another popular randomization and control group plan is the Solomon plan. The Solomon plan is an extended version of the RT1-2C plan that allows you to monitor and evaluate the effects of natural development and background, as well as determine the interaction of the test effect with the main exposure X. Here, along with the experimental and control groups with preliminary testing, experimental and control groups without preliminary testing are used (as in the RTC plan). Schematically, it looks like this:

Obviously, if the main effect X is real, then even if there is a significant testing effect (), four inequalities will hold: O2> O1; O2> O4; O5> O6; O5> O3. The value of the difference O6? O3 can serve as an estimate of the comparative magnitude of the effect of preliminary testing (without interaction with X). Comparison of O6c O1 and O3 allows us to assess the influence of background factors and factors of natural development.

So far, we have discussed the substantive aspects of testing the experimental hypothesis of a main effect X using different experimental designs, as well as the benefits of different designs for providing different aspects of validity. It is obvious, however, that in each specific experiment the magnitude of the main effect, i.e., the observed difference between the results of the experimental and control groups, will vary not only under the influence of the independent variable, but also simply as a result of the action of various random perturbations. Of course, if our experiment were ideal and absolutely reliable (see above), then for any number of repetitions we would always get the same true estimate of the magnitude of the impact (in the absence or control of the mixing effect of additional variables). However, real-life experiments - especially, as already mentioned, experiments in the social sciences - are not and cannot be perfect and irreproachably reliable. Consequently, the researcher is always faced with the task of statistically assessing the significance of the results obtained.

A probabilistic meaningful experimental hypothesis is a statistical hypothesis. Acceptance or rejection of a statistical hypothesis is a necessary but insufficient condition for the acceptance or rejection of a meaningful hypothesis being tested in an experiment. A statistical hypothesis tested in a particular experiment is always formulated as the hypothesis that with an infinite number of repetitions of this experiment, the average difference between the experimental and control groups (or between the effects of different levels of the independent variable) would be equal to zero. Such a statistical hypothesis, which actually amounts to a statement about the random nature of the differences observed in a real experiment, is called the null hypothesis, or the null hypothesis (H0). Rejection or non-rejection of the null hypothesis allows us to say that in this experiment the meaningful hypothesis was confirmed, or the opposite alternative hypothesis was confirmed, or none of them was confirmed.

Consider Fischer's imaginary experiment with tea and milk, described at the beginning of the chapter. We noted that using the tactics of random guessing, the subject could correctly determine the sequence of filling the cup in about 1 case in 70. Therefore, 2 out of 100, even with a very large number of trials, can hardly be considered a significant result. 90 or even 95 out of 100 look much more reliable (the rest can be attributed to the action of random factors). To determine the level of significance and build a confidence interval in this case, the method described in Chapter 8 is sufficient.

In general, the significance tests and statistical methods used to test a statistical hypothesis for a particular experimental design are called statistical models. For designs with a control group, the primary statistical model is the use of a t-test, as discussed in more detail below. For more complex designs of multivariate or factorial experiments, a general overview of which is given in the next section of the chapter, the leading statistical models are ANOVA and the use of Fisher's F-test.

So, in order to assess the statistical significance in the elementary randomized designs described above, we need to test the statistical hypothesis about the difference in the mean values of the dependent variable in the control and experimental groups. The specific value of the difference between the mean values of the dependent variable in the experimental and control groups, found in a separate experiment (say, 4 points for some), must be correlated with a certain interval in which this value with a given (confidence) probability. In other words, it is necessary to solve the problem of interval estimation, similar to the problem of estimating an individual parameter of the population in a sample survey (this problem is described in Chapter 8). The difference is that when conducting an experiment, we are interested not in the probable limits within which a certain characteristic of a sample from a really existing population lies, but in the limits within which the result we obtained in the experiment lies relative to the result of an imaginary infinite set of identical experiments. The null hypothesis asserts that the true value of the difference in means is equal to zero, varying within some limits from experiment to experiment (i.e., if it is possible to show that the value of the difference of group means obtained in the experiment does not allow accepting the null hypothesis, then it is concluded that confirmation of the hypothesis opposite to the zero (i.e., about the statistical significance of the differences between the groups - and, therefore, about the confirmation of the experimental hypothesis (or about the confirmation of the hypothesis opposite to the experimental one - if the difference between the experimental and control groups turned out to be with the opposite sign). that the null hypothesis is always formulated as the hypothesis that the true value of the difference between the means (or, say, the magnitude of the relationship between two variables) is zero, and the values obtained in the experiment differ from zero solely due to a random sampling error. - in one direction or another - the observed value is located, the more it is statistical significance and less likely to be the result of sampling error.

In order to compare the value of the difference between the means obtained in the experiment with the control and experimental groups with a hypothetical sample distribution of this value for an infinite number of trials (such distributions are available not only for the difference in means, but also for means, standard deviations, etc.) , you need to calculate the standard error of the difference between the means. Formula for standard error the difference between the means - SMe-Mk - differs slightly from the formula for the standard error of the mean SM20 given in Chapter 8. Nevertheless, it is quite simple:

where Se and SK are the standard deviation values calculated for the experimental and control groups, ne and nk are the number of observations (subjects) in the experimental and control groups.

After that, it is necessary to determine how many units of standard error are the obtained difference of means from zero, which is - in accordance with the null hypothesis - the mean of the hypothetical distribution of the differences of means, t-distribution. To do this, the difference between the group means obtained in the experiment must be converted into t-units (that is, the standard deviation units for the t-distribution). For a given difference in averages, the value of t can be calculated by the formula:

The obtained t value must be compared with the corresponding value from the t-distribution table for the chosen significance level (p = 0.05 or 0.01) and the number of degrees of freedom corresponding to the number of observations in each group (or subsample). The number of degrees of freedom is a rather complex statistical concept, the analysis of which is beyond the scope of this textbook (in the very general view it is discussed in Ch. 7). In practice, the number of degrees of freedom can be considered as a value equal to the number of observations (subjects, respondents, points, etc.) minus the number of parameters being evaluated. For the difference between the mean of the two groups, this will be the number of observations in the experimental group minus one (ne? 1) plus the number of observations in the control group minus one (nk? 1):

N article free = (ne? 1) + (пk? 1)

You can find t-distribution tables in any textbook or reference book on statistics (see this chapter and also chapter 8). Here we present only a fragment of such a table.

Table 4.1

Abbreviated t-distribution table

Student (W. Gosset, 1908)

Number of degrees of freedom

Consider an example of calculating t for the experiment described above, which examined the effect of an anti-war film on changing student attitudes. Let the following results be obtained for the control and experimental groups during the final testing on the scale of pacifist attitudes: Control group Experimental group

Ne = 34 people

Our statistical challenge is to determine if the means of the two groups differ enough to reject the null hypothesis that the means are from the same population. Let's use the above formula to calculate the value of t:

The number of degrees of freedom in the given example: (28 × 1) + (34 × 1) = 60.

The obtained value t = 3.4760 obviously exceeds the tabular values and for p

It is important, however, to always remember that the statistical significance of the results is completely different from their meaningful significance! Even the high statistical significance of the experimental results does not guarantee that these results will have any interesting interpretation and will affect the state of modern sociological knowledge. Substantive significance depends primarily on our ability to link experimental hypothesis with existing sociological theories.

Multivariate and factorial experiments:

general review

In the experiments described above with the control group, only two types of conditions are used each time - either. These two types of conditions can essentially be thought of as two levels of the independent variable that can be assigned conditional numeric values such as and. In other words, from the point of view of the measurement level, the independent variable is nominal, qualitative. In the control group, its value is equal to zero, in the experimental group - to one. However, the researcher often has much more information about the independent variable and is able to measure and control it at at least three to four levels of values. Accordingly, the experimental hypothesis can be formulated in terms of more or less intense exposure or presence-absence of the dependent variable at specific levels of the independent variable.

In psychology, there is a well-known law, the so-called Yerkes-Dodson law.

At the beginning of this century, R. Yerkes studied how negative reinforcement in the form of an electric shock affects the development of elementary skills in animals. In particular, in experiments with (a type of domestic mouse that has a genetic defect that makes it move in a circle or in a figure eight), he used three levels of current strength - (500 conventional units), (300 conventional units) and (125 units). The mouse had to learn to choose one of two tunnels. At the end of the tunnel, a mouse of the opposite sex was waiting for her anyway. With the wrong choice (white tunnel), the mouse experienced an electric shock; with the right choice (black tunnel), there was no negative reinforcement. The location of the tunnels (left-right) varied randomly from sample to sample. It turned out that the fastest learning occurs when the amount of stimulation. The nonlinear nature of the relationship between the magnitude of the incentive to solve a certain problem and the success of the solution discovered in this experiment was then repeatedly confirmed in many other experiments, including with human subjects and with positive stimulation. Excessive motivation and excessive amount of reinforcement, as well as weak motivation, each time had a smaller impact on the success of various tasks.

Experiments that use several (more than two) levels of the independent variable are called multilevel experiments. The scheme of the above experiment with randomization and three levels of the independent variable (X1X2, X3) is as follows: R X1 O1

The experimental hypothesis in this case is formulated as a hypothesis about the ratios of the values of O1, O2, and O3 (in the considered example, O1 O3). An independent variable in a multivariate experiment can have more than three levels. In other words, it can be a quantitative variable, measured at an interval or absolute level. Accordingly, the hypothesis of a multidimensional experiment can be formulated in more precise terms - as a hypothesis about or even relations of variables. For example, the experiment can study the influence of the attractiveness of the lecturer on the frequency of attending classes by students, the effect of the number of available sources of information about the product on the formation of consumer preferences, or the nature of the relationship between the amount of monetary remuneration of the subjects and the success of solving similar problems. Thus, multidimensional experiments allow testing more subtle and accurate meaningful hypotheses about the mechanisms of individual and group behavior.

Statistical hypotheses tested in multivariate experiments are hypotheses about the differences between the values of the dependent variable for different levels of the independent variable. The null hypothesis is formulated as a hypothesis that the scatter of individual values within one level of the independent variable (within the corresponding experimental group) is identical to the scatter of individual values between different levels (groups), i.e., the ratio of the variance of intergroup estimates to the variance of intragroup estimates is 1. The last the ratio is referred to as the F-criterion. In order to determine whether the value F obtained in a particular 8 experiment exceeds the threshold value of the statistical F-distribution for a given level of significance, the statistical technique of one-way analysis of variance is used. The term in this case means that only one independent variable (influence factor) was used in the experiment. Consideration of the technique of analysis of variance and statistical estimation of the resulting F value is beyond the scope of this review (detailed descriptions and recommendations, if necessary, can be found in books from the list of additional literature to the chapter).

In those areas of sociology and social psychology that have a relatively developed tradition of experimental research (interpersonal and intergroup perception, studies of the dynamics of attitudes, social processes in small groups, evaluating the effectiveness of educational programs, etc.), more complex experimental schemes are often used, united by the term ...

A factorial experimental design includes two or more independent variables (also referred to as), each of which has several levels of impact. Since with an increase in the number of independent variables, the number of groups grows very quickly, in each of which one of the possible combinations of these variables and their levels is used (in the full factorial plan, the number of groups is equal to the product of the number of levels set for each independent variable), in order to save resources and rational distribution of research efforts, numerous plans were developed, where each of the variables is implemented once, and the generalization and statistical analysis of the interaction of various factors and their isolated and joint influence on the dependent variable is carried out at the group level.

Any factorial experiment is, in essence, several experiments combined in one plan. The generalized data of the factorial experiment allow us to answer two types of questions: 1) whether there is an influence effect for each separately taken independent variable; 2) does the magnitude of this effect of influence depend on the magnitude of the values of other independent variables? The isolated effect of one independent variable is called the main effect, and the change in the magnitude of this effect under the influence of another independent variable is called interaction.

Table 4.2 presents the design of a simple factorial experiment (), in which the effect of novelty and type of image on the interest shown in this image by 4-month-old infants is presented. The gaze duration was used as an indicator of interest. Each of the independent variables was represented by only two levels: for novelty - a new or old image presented in the previous series; for the type of image - a geometric contour or a schematic representation of a human face (schematic drawings were used to equalize images in terms of visual complexity, since the time for fixing the gaze usually depends on the complexity and number of details). As can be seen from the data in Table 4.2, both main effects are evident. The influence of novelty on interest becomes obvious when comparing the averages for the rows - the average duration of looking at images (both geometric and) is noticeably higher in the case of presentation of new drawings (55 seconds versus 20). Comparison by columns shows that when the data are averaged over two groups (new and old drawings), the images of a human face evoke much more interest, which is manifested in a longer gaze (45 sec). There is also an interaction between image type and novelty. Presentation results different types images are different for and group. The values of the differences by columns for each row also differ (60 -50 = 10 compared to 30 -10 = 20), and the corresponding indicators by rows (60 -30 = 30 compared to 50 -10 = 40). In other words, the greater attractiveness of human faces is more pronounced when old drawings are presented (the difference of 10 seconds when new pictures are presented increases to 20 for old images), and the difference between the presentation of old and new drawings when using geometric contours increased to 40 seconds.

Table 4.2

Factor Experiment 2x2

The novelty of the image

Image type

Average gazing duration, sec.

Geometric

Average duration, sec.

When processing the results of multivariate experiments, the main statistical model is multivariate analysis of variance.

Multivariate experiments in sociology are very often field experiments that simulate complex relationships in the real world. The advantage of field multifactorial experiments lies in their, i.e., external, facial validity. But here lies the main disadvantage of such experiments - lower reliability and internal validity. Critics of field multifactorial experiments often note that the approximation of an experiment to the real world is often achieved here by replacing the experimental control with a purely statistical one. In the latter case, threats to validity associated with incorrect specification of the measurement model increase from individual levels of independent variables with uncontrolled external variables (see Chap. 5, 6). In addition, in multifactorial experiments, the problem of data aggregation is more acute than in individual and intergroup experiments - there is almost always a possibility that the relationships revealed in the analysis of summary group data are not exactly observed for any individual subject (just like the mean of a certain sample may not apply to any specific sample observation). various effects interaction of variables-.

In the social sciences, the concept of quasi-experiment, or quasi-experimental research plan, is also often used. We are talking about panel, trend, etc. plans for sample surveys (Chapter 5). Sample surveys, especially ongoing or conducted as comparative studies for subsamples that have experienced or have not experienced a specific, localized impact (for example, a social revolution, educational reform, or a stock market crash), do indeed allow one to draw conclusions about the relationship between the independent and dependent variables of interest to the researcher. , and therefore - to test hypotheses about the alleged causal relationships, however, experimental randomization and control in sample studies, as shown in the relevant chapters, are replaced here by the use of random samples and special methods of statistical data analysis.

additional literature

Voznesenskiy VA Statistical methods of experiment planning in technical and economic research. 2nd ed., Rev. and add. Moscow: Finance and Statistics, 1981. Ch. 2, 3.

Glass J., Stanley J. Statistical methods in pedagogy and psychology: Trans. from English / Common. ed. Yu. P. Adler. M .: Progress, 1976. Ch. 15-19.

Gottsdanker R. Foundations of a psychological experiment: Per. from English M .: Publishing house of Moscow State University, 1982.

Druzhinin N.K.Selective observation and experiment. Moscow: Statistics, 1977.

Campbell D. Models of Experiments in Social Psychology and Applied Research: Per. from English / Comp. and total. ed. M. I. Bobneva. Moscow: Progress, 1980.

Yadov V.A. Sociological research: methodology, program, methods. M .: N

Experiment(from lat. experimentum -

Without any exaggeration, scientific observation is an "environmentally friendly" method of cognition. In it, a sparing regime is practiced in relation to the object of knowledge.

Experiment is another matter altogether. An artificially created situation is studied here, which either does not occur in nature at all, or does not occur in its pure form. If the experimental situation begins to resemble a natural one - that's it, put out the lights in the laboratory, start all over again. "

The experimental method should be attributed to the category of tough, not just contact methods of a sports duel, but, if you like, to the category of a real combat duel, where man and nature came together for life and death. It is no accident that they say: this is not an experiment for you, but real life, thereby emphasizing the dissimilarity of the experimental situation to the natural one.

The contact rigidity of the experiment, if such an expression is applicable here, pays off with interest. A scientist creates a situation that does not happen in nature, looking, as they say, into its inside out, and in this way establishes something that is impossible in any other method, namely, cause-and-effect relationships. In the experiment, we eliminate all the random features of the phenomenon, all the factors that distract, distort the result, and get to the very bottom. And the essence of any phenomenon is known to everyone - universal, natural, necessary.

Experiment should not be confused with at least three of the following processes:

♦ test;

♦ innovation;

These concepts have a lot in common with experiment. Indeed, in all cases, there is some change in the initial state of the object. In all cases, someone monitors how the change affects the state of the object and then records the results.

But there is a major difference. It is only in a scientific experiment that a research program is developed in advance, goals, objectives, subject and object of research are formulated, hypotheses are put forward, concepts are operationalized and methodological techniques for testing a hypothesis are developed. Only in a scientific experiment, a so-called experimental group is formed, over which planned manipulations are performed, for example, they teach students to read speed or make them sit at a computer for 12 hours, and a control group, which remains in its usual conditions and is not subject to any manipulations.

Experiment(from lat. experimentum - trial, experience) - a research strategy, during which purposeful observation is carried out of how one characteristics of a phenomenon (process, situation) behave, if other characteristics are changed sequentially and in a certain order.

3.2. Methodology and experimental technique

Although the methodology and technique of experiment came to sociology from modern psychology, the logic of the experimental method itself was developed by the English philosopher John Stuart Mill, who lived in the 19th century. He established several logic circuits for inductive inference, one of which - the method of distinction - is a classic experimental design. The general principles of the experiment are associated, firstly, with the formation of initial hypotheses about the cause-and-effect relationships between the properties (variables, factors) of the object under study, and secondly, with the forcing of a special artificial situation that allows you to purposefully influence the controlled variables in order to test the desired hypothesis about the presence (or absence) of relationships between controlled and uncontrolled variables. The result of the experiment is the acquisition of new knowledge about the object under study.

The main goal of the experimental method is to test certain hypotheses, the results of which have a direct impact on practice. As a kind of in-depth, analytical sociological research and at the same time as a method of collecting information about the factors affecting the change in the state of certain social phenomena and processes, as well as the degree and results of this impact, the experiment is of great scientific and practical value.

Experiment Subject - relations and connections between the properties of the studied object and the system of factors (variables) that determine its social activity and behavior.

Variables in the experiment - controlled and uncontrolled factors that have a direct or indirect effect on the state of the object under study. The experimental situation is specified and described by the system of variables. The main variables are independent and dependent. Independent variable is a factor changed by the researcher and introduced into the activity of the experimental group. This is a relatively independent, stable, significant factor influencing the state of the object. Dependent variable - a factor that changes under the influence of the independent variable. Variables represent analysis categories in an experiment and, like analysis categories, are displayed as empirical measures 2.

Experimental situation - a set of conditions in which the experiment takes place. They can be field and laboratory.

Experiment object - social group, team, personality (participants in the experiment). In the experiment, two types of objects (experimental units) are distinguished: control and experimental groups. Comparison of the two groups reveals the difference and allows you to judge whether the expected changes have occurred or not. The number of people participating in the experiment is usually small and rarely exceeds 10-15 people.

Experimental group - the one that is influenced by the experimental factor (independent variable). In addition to the general requirement for the choice of an object - representativeness in the class of objects under study - this group has requirements for the relative stability of the composition for the period of the experiment, the ability to conduct systematic controlled observation; in some cases, the consent of the group is required to participate in experiment 3.

Control group identical to the experimental one in terms of the parameters set by the researcher, but not experiencing the influence of experimental factors (independent variable). In the presence of a control group, the purity of the social experiment increases due to the fact that the effect of side parameters is equally likely in both the experimental and control groups.

3.3. Experimenting procedure

The procedure for conducting the experiment is that scientists first plan to divide the research object into two homogeneous groups and introduce several new factors into one of them, and leave the second unchanged. Then a comparison of the two groups is organized. The object with the help of which it is carried out often acts only as a means for creating an experimental situation. Therefore, in the future, unlike, for example, the object of a mass survey, it may not be relevant to the use of the results and conclusions.

The general logic of the experiment is that by choosing a certain experimental group (or groups) and placing it in an unusual experimental situation (under the influence of a certain factor), trace the direction, magnitude and stability of changes in the characteristics of interest to the researcher, which can be called control ones.

The instrumentation of the experiment is similar to the instrumentation of the observation method and includes: a protocol, a diary, cards, questionnaires are possible.

3.4. Forming groups in an experiment

An important condition for the experiment is the formation or selection of experimental groups.

Sociology employs three methods of group formation: pairwise selection, structural identification, and random selection 5.

Pairwise selection method: from the general population, two groups are selected so that they are identical in terms of neutral and control, but differ in factorial characteristics 6. After that, the same conditions are created for both groups, and after a while the effect of the experiment is measured by fixing and comparing the parameters of control signs in both groups. The method is used mainly in a parallel experiment.

Structural identification method used in both linear and parallel experiments. In a linear experiment, a group is selected so that it is a macromodel of the general population in terms of neutral and control characteristics.

Random sampling method Identical to the previously discussed preset size probabilistic sampling methods. As a rule, it is used in field experiments with a large (up to several hundred people) size of the experimental group 7.

Randomization is widely used in the technology of random selection. Randomization (randomisation) - a procedure that provides a random distribution of subjects in the experimental and control groups. The goal is to minimize the possibility of conscious or unconscious selection of subjects of the same type, for example, the most educated or the healthiest.

3.6. Experiment classification

In the social sciences, experiments are subdivided into:

Laboratory experiment- a methodological strategy aimed at modeling the activity of an individual in special conditions. It is based on the creation of artificial situations that allow you to change the controlled variables and monitor the accompanying changes in the dependent variables. Most often, this method requires special equipment, and sometimes the use of technical devices. The leading feature of a laboratory experiment is to ensure the reproducibility of the studied characteristic and the conditions for its manifestation. At the same time, the requirement for the purity of the data obtained is associated with some limitations: in artificial laboratory conditions, it is almost impossible to simulate real life circumstances, only their individual fragments.

The laboratory experiment promotes a deep and comprehensive study of the mental activity of people. In laboratory conditions, social-psychological and psychological experiments are usually carried out.

The Stanford Prison Experiment is a famous psychological experiment that was conducted in 1971 by the American psychologist Philip Zimbardo. The experiment is a psychological study of a person's reaction to the restriction of freedom, to the conditions of prison life, and to the influence of the imposed social role on behavior.

If you give a person power over someone helpless, someone humiliated, then absolute power corrupts absolutely.

Professor David Wilson, criminologist

Volunteers played the roles of guards and prisoners and lived in a conditional prison, arranged in the building of the Department of Psychology. Prisoners and guards quickly adapted to their roles, and, contrary to expectations, truly dangerous situations began to arise. Every third guard showed sadistic inclinations, and the prisoners were severely traumatized, and two were prematurely excluded from the experiment. Despite the apparent loss of control over the experiment, only one in 50 observers, Christina Maslach, opposed its continuation. Zimbardo ended the experiment ahead of time.

The results of the experiment were used to demonstrate the receptivity and submissiveness of people when there is a justifying ideology supported by society and the state. They were also used as an illustration to the theory of cognitive dissonance and the influence of the power of authorities.

Field experiment takes place in natural conditions for the process under study. The technique of this experiment requires the selection of control and experimental groups with the necessary set of variables. In this case, the members of the group (subjects) may or may not be aware that they are participating in the experiment. The decision to inform them in each specific case depends on how much this awareness can affect the course of the experiment.

Field experiments using a simplified scheme are often carried out by journalists. For a while, disguised as representatives of a social group or profession, for example, beggars, they go out into the street, go down the subway or enter public transport, begging for alms. the main objective- see the reaction of people. They get excellent first-hand information. The result is a journalistic sketch of real events in vivid colors.

True experiments organized in compliance with all the requirements of the scientific method developed by natural science (research program, hypotheses about cause-and-effect relationships, dependent and independent variables, statistical selection of experimental and control groups, scientists do not control all factors that affect the final result , etc.) are quite rare in the social sciences. Any rule is necessarily violated or not fully complied with. Justifying themselves, social scientists point to the complexity of the subject matter of their science, the interference of the human factor, and many other reasons that are absent, say, in physics.

Quasi-experimentation is much more commonly used in the social sciences. It is considered to be the opposite of true experiment 15. In quasi-experiments, all conceivable and unimaginable requirements of the scientific method are violated, namely, there is no research program, no hypotheses and variables are put forward, two mandatory types of group are not created, etc., nevertheless, the experiment is considered to have been carried out 16. From an experiment in such a procedure, sometimes one thing remains - the effect of the subject on the object, a change in the social environment.

Natural experiment very different from true and quasi-experiments. In the second case, any intervention is arranged by the scientist, in the first it occurs naturally. Natural cases include the following: a) some of the residents decided to leave the village for the city, and some stay; b) some villages in the region were supplied with electricity, while others were not; c) part of the representatives of the national minority attends schools for the representatives of the middle class, and part of them study on the reservation. Any of these situations can become the object of a natural experiment, during which aspects of human behavior are studied. In such cases, it is impossible to pretest, measure the independent variables before the intervention. The scientist, theoretically or from secondary sources, mentally reconstructs the initial conditions, then studies the course of the experiment and the consequences. Often he only finds the latter, and everything else has to be reconstructed according to the polls of the respondents.

Linear experiment is carried out without control groups and is based on a comparison of the state of controlled and dependent variables before and after experimental exposure. It differs in that the analysis is carried out on the same group, which is both the control group (its initial state) and the experimental one (its state after changing one or several characteristics). In other words, even before the start of the experiment, all the control, factorial and neutral characteristics of the object are clearly recorded. After that, the factor characteristics of the group (or the conditions of its functioning) change, and after a certain predetermined time has elapsed, the state of the object is again measured according to its control characteristics. It is very important that in the process of a linear experiment the influence of interfering factors on the object of analysis is excluded.

Parallel experiment involves the use of control groups, in which the state of independent and dependent variables is measured in parallel with their state in the experimental group, and experimental influences are not performed in the control group. The composition of the groups should be identical for all controls, as well as for neutral characteristics that can affect the outcome of the experiment (first of all, these are socio-demographic characteristics). The characteristics of the control group remain constant throughout the entire period of the experiment, while the characteristics of the experimental group change. Based on the results of the experiment, the control characteristics of the two groups are compared and a conclusion is made about the causes and magnitude of the changes that have occurred.

Depending on the activity of the experimenter, there are active and passive experiments. The first provides for a purposeful change in the studied situation by introducing a controlled variable, which is a hypothetical cause of the state of the dependent variable. The experimenter resorts to managerial influence on the object under study: a new form of remuneration is introduced, a sliding schedule of days off, a quality control system, etc. A passive experiment is characterized by the selection in natural conditions of objects possessing the combination of variables necessary for the experimenter.

Natural experiment- an experimental strategy developed by A.F. Lazursky in 1910. It is characterized by the fact that it is carried out under conditions close to the usual activity of the subject, who does not know that he is participating in the research. Due to this, a greater purity of the experiment is achieved. The main methods are observation and conversation with the subject, the results of which are processed qualitatively. Such an experiment combines the positive aspects of the observation method and the laboratory experiment. Here, the naturalness of the observation conditions is preserved and the accuracy of the experiment is introduced.

Learning experiment - a form of natural experiment, which is characterized by the fact that the study of certain mental processes occurs during their purposeful formation. With the help of this method, it is not so much the present state of knowledge, abilities, and skills that is revealed, but the peculiarities of their formation.

Within its framework, the subject is first asked to independently master a new action or new knowledge (for example, to formulate a pattern), then, if this did not succeed, he is provided with strictly regulated and individualized assistance. This whole process is accompanied by an ascertaining experiment, thanks to which it is possible to establish the difference between the initial, actual level and the final one corresponding to the zone of proximal development.

Manufacturing experiment- a natural experiment carried out in the usual conditions of the subject of his work. At the same time, the employee himself may not know about the conduct of a production experiment. In other cases, he becomes an active participant in the experiment, which is important, for example, when the structure of labor activity changes.

Experiment as a form of religious observer's work in modern media Coursework >> Journalism

Qualitative methods is sociological experiment. This method most ... connections. (18.) sociological experiment may be separate ... .- M., 1979 6. Devyatko I.F. Methods sociological research.- Yekaterinburg, 1998 Dzyaloshinsky I.M. ...

Features of social experiments

Report >> Sociology

Rather, a principled character. If, for example, in sociological experiment the independent variables are race or ... an experimental hypothesis with existing sociological theories. Multidimensional and factorial experiments: general overview In the described ...

Theory and methodology sociological research

Book >> Sociology

Experimental impact. Usually the result of uncontrollable sociological experiments with a naturally occurring independent variable ... series and control group. Many sociological experiments are built according to a simplified plan "research ...

Ministry of Education and Science of the Russian Federation

Federal State Budgetary educational institution

higher professional education

"Tyumen State Oil and Gas University"

Department of Sociology and Political Science

TEST

In the discipline "Sociology"

On the topic: "Experiment in Sociology"

Introduction

The concept of experiment in sociology. Classification. Types

2. Technique of the experiment

Equalization of experimental conditions

Experiment program and instrumentation

List of used literature

Introduction

Experiment is a specific method based on the controlled interaction of the researcher with the investigated object under predetermined conditions. In an experiment, information can be obtained in an artificially created environment, which distinguishes this method from ordinary observation.

Various social groups act as objects of sociological experiments.<#"justify">1. The concept of experiment in sociology

Experiment in sociology is a method of collecting and analyzing empirical data aimed at testing hypotheses about causal relationships between social phenomena. In a real experiment, this check is carried out by the experimenter's intervention in the natural course of events: he creates or finds a certain situation, activates a hypothetical cause and observes changes in the situation, fixes their correspondence or inconsistency with the hypothesis put forward.

A hypothesis is an assumed model of the phenomenon under consideration. On the basis of this model, the phenomenon under study is described as a system of variables, among which an independent variable (experimental factor) stands out, subordinate to the control of the experimenter and serving as a hypothetical cause of certain changes in the dependent variable. Non-experimental variables are properties and relationships that are essential for the phenomenon under study, but since their influence is not investigated in this experiment, they must be neutralized (isolated or set constant).

The main features of a social experiment are:

active intervention of the researcher in the system of the studied phenomena;

systematic introduction of a relatively isolated experimental factor, its variation, possible combination with other factors;

systematic control over all significant determining factors;

the effects of changing dependent variables should be measured and unambiguously reduced to the influence of independent variables (experimental factor).

The structure of a social experiment can be represented as follows:

Experimenter. This is usually the researcher or group of researchers who design and conduct the experiment.

An experimental factor (or independent variable) is a condition or system of conditions that is introduced by a sociologist. The independent variable must be, first, controllable, i.e. its direction and intensity of action should be in accordance with the program settings; secondly, controlled, if its quality and quantitative characteristics are disclosed within the experiment program.

Experimental situation - a situation that is created in accordance with the research program for the experiment. The experimental factor is not included in the conditions of the experimental situation.

An experimental subject is a group of individuals who have agreed to participate in an experimental study.

Classification of sociological experiments

Experiments carried out in sociology differ: 1) by the nature of the object and subject of research; 2) according to the specifics of the task at hand; 3) by the nature of the experimental situation; 4) according to the logical structure of the proof of the hypothesis.

Table 1 Classification of sociological experiments

Basis for classification Types of experiments The nature of the object and subject of research Real (natural) on the general population on the sample population Thought on a mathematical model "ex-post factum" experiments Specificity of the problem posed Scientific: theoretical and methodical Applied Projective Retrospective One-factor Multifactorial Characteristic experimental situations

A natural (or field) experiment can be controlled and uncontrolled. In uncontrolled experiments, the result is achieved by sufficient a large number repeated experiments, so that, according to the theory of probability, uncontrollable factors mutually cancel out and would not affect the effect of the experimental factor.

More rigorous data can be obtained in a controlled field experiment.

A controlled (valid) experiment is an attempt to obtain a relatively pure effect of an experimental variable. For this purpose, a careful alignment of other conditions is undertaken, which can distort the result of the influence of the experimental factor.

A sociological experiment is fundamentally different from a natural science experiment. The peculiarity of the latter is that the object is the material world, explored with the help of a certain device or instrument, i.e. the experimenter, in the words of G. Hegel, "acts against nature with the help of nature itself", while a sociological experiment is cooperative activity subjects and a sociologist, aimed at researching any particular feature of a person, a group.

This method is used to test hypotheses regarding causal relationships between social phenomena. At the same time, two complex phenomena are compared, differing in that in the first there is some hypothetical cause, and in the second it is absent. If, under the influence of the experimenter, a change is observed in the first, but not in the second, then the hypothesis is considered proven. Experimental research in sociology differs from the methods of other sciences in that the experimenter actively manipulates the independent variable. If in the use of non-experimental methods, as a rule, all groups are equal for the researcher, then the experiment usually involves the main and control groups of subjects.

Due to the different level of development of a particular scientific problem and the lack of information about the relationship between the dependent and independent variables, two main types of experiments are distinguished:

§ research, which is carried out in the case when the causal relationship between the dependent and independent variables is unclear and the experiment is aimed at testing the hypothesis of a causal relationship between two phenomena;

§ confirming, which is carried out if the connection is clarified in advance and a hypothesis is put forward about the content of the connection. Then, in the experiment, this connection is revealed and refined.

So, when clarifying the causes of social tension in a particular city, the following are put forward possible hypotheses: low incomes of the population, social polarization, unprofessional administration, corruption, negative impact Media, etc. Each of them requires verification, although it seems quite reasonable.

The experimenter must have the necessary information on the problem under study. After the formulation of the problem, the key concepts are determined that are contained in the special scientific literature and sociological dictionaries. When working with literature, not only the problem is clarified, but also a research plan is outlined, new hypotheses arise. Next, the variables are defined in terms of the experimental procedure; first of all, external variables are highlighted that can significantly affect the dependent variable.

The selection of subjects must meet the requirement of representativeness, i.e. be made taking into account the characteristics of the general population, in other words, the composition of the experimental group should model this population, since the conclusions obtained as a result of experiments apply to the population as a whole.

In addition, the subjects should be assigned to the experimental and control subgroups so that they are equivalent.

The researcher experimentally influences the first group, but in the control group there is no effect. As a result, the resulting difference can be attributed to the independent variable.

Suppose a researcher hypothesizes that in a given city, the influence of the media leads to an increase in social tension. But what is the cause and what is the effect? It is possible that social tension itself influences the nature of television broadcasts and the publication of “disturbing” articles in the local press. In this case, a sociologist can conduct an experiment to find out this causal relationship.

So, for the experimental group, it is possible to control (decrease or increase) the number of programs with excessive "negative" information, change the influencing factors in order to find out how these factors separately or collectively affect people, i.e. the researcher manipulates one or two independent variables, trying to keep all the others unchanged (Fig. 1).

Thus, a sociological experiment should be understood as a method of collecting and analyzing data that makes it possible to test hypotheses about the presence or absence of causal links between social phenomena. To do this, the researcher actively intervenes in the natural course of events: creates in the studied group artificial conditions and systematically controls them.

Rice. 1. The impact of the media on the growth of social tension

The information obtained during the experiment about the change in the indicators of the studied object contributes to the clarification, refutation or confirmation of the initial research hypothesis. The experimental method allows obtaining reliable results that can be successfully applied in practice, for example, to increase the efficiency of the functioning of social groups, organizations, institutions. However, in the process of applying the experimental method, it is important to take into account not only the reliability of the data, but also the moral and legal regulations as well as the interests and aspirations of the people involved in the study.

Usually (in a real experiment) this check is carried out by the experimenter's intervention in the natural course of events: he creates or searches for a certain situation, puts into action hypothetically. The reason and observes changes in the situation, fixes their compliance or inconsistency with assumptions. Experiment as research and transformation activity<#"justify">1)By the nature of the object and subject of research;

)By the specifics of the task at hand;

)By the nature of the experimental situation;

)According to the logical structure of the proof of the hypothesis.

By the nature of the object of research, real and mental experiments are distinguished. A real experiment is characterized by purposeful intervention in reality, testing explanatory hypotheses by systematically changing the conditions of social services. activities. In a thought experiment, hypotheses are tested by unrealistic phenomena, and information<#"justify">2. Technique of the experiment

One of the well-known methods of collecting primary information in sociology is a social experiment. The main task of its application is the need for practical testing of theoretically established causal laws. At the same time, the experiment should be viewed as a method for testing a scientific hypothesis. In this case, the goal of the social experiment is to identify a stable real pattern in the form of a clearly expressed cause-and-effect interdependence of the signs of social objects.

The most complete definition of an experiment is given in the "Sociologist's Workbook": "An experiment in sociology is a way of obtaining information about the quantitative and qualitative change in the exponential activity and behavior of an object as a result of the impact on it of some controlled and controlled factors."

The need to apply a social experiment arises when it is necessary to solve problems associated with the reaction of any social group to the introduced internal or external factors, if these problems cannot be solved in another way.

The most important thing in a social experiment is the creation of appropriate conditions (meaning the optimal choice of experimental factors). The factors (characteristics) that are most important from the point of view of the problem under study, the creation of an experimental situation are selected as control factors. In turn, the change in the control characteristics should depend on those characteristics of the experimental group that are introduced or changed by the researcher himself. Such characteristics are called factorial. Characteristics that do not matter from the point of view of solving research problems are called neutral.

Types of experiment and methods of selection of experimental groups.

Experiments differ both in the nature of the experimental situation and in their logical structure.

By the nature of the experimental situation, experiments are divided into field (the object is in familiar, natural conditions) and laboratory (artificial formation of groups, placing them in specially created conditions),

According to the logical structure of proving hypotheses, linear and parallel experiments are distinguished.

In a linear experiment, the same group is analyzed, which is both the control (in the initial state) and the experimental (in the state after the experiment).

V parallelthe experiment involves two groups at the same time: control and experimental. Their composition should be the same in control and neutral characteristics, moreover, the characteristics in the control group remain constant, and in the experimental group they change. At the end of the experiment, the control characteristics of the two groups are compared. Then conclusions are made about the reasons and extent of positive or negative changes in the research object.

The very logic of building and implementing a social experiment can be represented by the following structure:

Determination of the purpose of the experiment.

Object selection (control and experimental groups).

Definition of the subject of research.

The choice of control, factorial and neutral signs.

Determination of experimental conditions and creation of an experimental group.

Defining tasks.

The choice of methods to control the course of the experiment.

Determination of methods for testing the effectiveness of the experiment.

The logic of the experiment is always subordinated to the search for the reasons, the nature of the change of the social phenomenon or process of interest to the researcher.

The typology of social experiments is presented in Table 1.

Table 1 Types of social experiments

Major divisions Types I. Specificity of tasks. 1. Scientific research; 2.Practical. II. The nature of the experimental situation. 1. Field; 2.Laboratory.III. The logical structure of the proof of the hypothesis. 1. Parallel; 2. Consecutive. IV. The nature of the object. 1. Social; 2.Pedagogical; 3.Economic, etc.

Let's give an example of a social experiment in the field of physical culture and sports.

The purpose of the experiment is to reveal the influence of different styles of students' activity management (authoritarian and democratic) in physical education lessons on their physical activity. The control and experimental groups were organized taking into account the purpose and objectives of the study. The subject of the research is the effectiveness of physical education classes in conditions of different styles of students' activity management. The control characteristics in this study are the level of students' motor activity in the classroom, factorial - different styles of leadership. The quality of execution can be neutral. motor actions... The experimental situation is set by conducting classes in the control and experimental groups different styles manuals.

The hypothesis underlying this social experiment was the assumption that the use of democratic style management of students' activities in physical education classes, providing for some freedom of choice of types of motor activity, will lead to an increase in motor activity.

In this case, the hypothesis was tested by determining the motor density of classes.

In conclusion, we note that modern practice shows that experimental research are a reliable method of collecting information and scientific analysis.

Equalization of experimental conditions

Equalization of conditions applies to all objects participating in experiments: experimental and control. Experiments are also possible without a control object, repeated several times. Then the conditions of the experimental objects in each series of experiments are subject to equalization.

Before proceeding with the equalization of conditions, it is necessary to highlight the characteristics that presumably affect the expected effect. It requires preliminary analysis problems in developing a research program. All of these characteristics are potentially experimental variables. But in each individual experiment, the effect of only one of the selected factors is checked, and the rest are subject to equalization.

First of all, the main parameters of the general social situation are leveled, such as the type of settlement, industry, ethnic and cultural environment, time interval and other features that are equally applicable to all objects of the studied process.

The main techniques for aligning individual characteristics in the case when the observation units are individuals are as follows.

Point alignment is used in small group experiments. The procedure is reduced to the selection of individuals in the groups to be aligned according to common features, identified as essential, such as, for example: a) age; b) marital status; c) gender; d) permanent residence, etc. Then, when aligning in the experimental and control groups, each individual in the first group should be found an analogue in the second: Ivanov, a third-year student, 20 years old, unmarried, nonresident should correspond to Petrov with exactly the same data. This method is effective for small groups, therefore it is used, as a rule, in laboratory experiments and extremely rarely in field research.

Frequency alignment involves comparing essential features in the experimental and control groups in proportions, mean values, etc., in accordance with similar indicators of the general population (see Table 2) ..

In this case, there is a danger that there may be different combination values of the selected signs in the experimental and control groups, which also differs from their distribution in the general population. For example, if, for example, one group of law and accounting faculties are taken as the experimental and control groups, then they may have a different ratio of men and women, or urban and nonresident, which differs from the distribution of these parameters in the entire general population, i.e. ... among all students of the academy.

Table 2 Frequency equalization of individual characteristics in a controlled experiment (in%)

Characteristic to be leveled General population (students of the academy) Group Experimental Control Gender Male 424040 Female 586060 Residence Urban 757575 Foreign country 252525

Quota alignment, often used in large samples, eliminates the disadvantages of the previous method. In this case, the groups are compared according to the proportional representation of characteristics taken in rigid combinations (quotas), as shown in table. 3

Table 3 Equalization of individual characteristics in a controlled experiment by the quota method (in%)

Quota of characteristics to be leveled Experimental group Control Men - third-year lawyers, urban, age 20 years 1010 Women - third-year lawyers, urban, age 20 years 1515

Random mechanical alignment is used in mass experiments, on large objects, when individuals are selected according to the rules of random non-repeated sampling. However, this technique is not suitable for small groups.

A kind of controlled field experiments - randomization (from the English random - randomness), which allows you to eliminate or minimize the impact of uncontrolled "random" factors.

Using a significant number of experimental and control objects or conducting numerous experiments, randomization allows you to "extinguish" the influence of uncontrolled (background) influences, if they are not systematic. Then the experimental effect is assessed by the usual calculus of the significance of the means according to the criteria of the state "before - after" on the experimental and control objects, and the degree of significance of the revealed regularities is determined using correlation analysis.

A thought experiment, or a quasi-experiment, can be a rather fruitful method of sociological research. Its peculiarity lies in the fact that instead of manipulating real objects, we operate with information about the events that have taken place.

In a full-scale experiment, the researcher designs the alleged events, acting on the experimental object with their hypothetical causes. In a mental experiment, the line of reasoning is the opposite: from the available consequences to possible reasons... This experimental move is called a retrospective analysis or an "ex-post-facto" experiment.

For example, you can retrospectively check the assumption that the nature of a student's creative activity significantly affects the directions of his subsequent career (remember Zhvanetsky: "Ilchenko and Kartsev and I graduated from the Odessa Institute of Engineers water transport... It is clear what kind of institute it is, if you look at what specialists it graduates. ") In this case, we check to what extent the student's participation in scientific research determines his future career in the field of science. scientific field - defended a dissertation, works at a university or research institute, etc. Then we check their presence in the lists of participants in the research work in the years when they studied at the institute. on average for the contingent of university students, then, obviously, the assumption made is justified.

Experiment program and instrumentation

An experiment program is a description of an experimentally tested hypothesis and procedures for testing it (system of variables, experimental factor, experimental situation (conditions), experimental and control groups, experimental instrumentation).

The experimental toolkit includes a protocol, diary and observation card.

The main resulting document of the experimental method is the protocol of the experiment, in which the following positions:

The name of the topic of the experiment.

The exact time and place of its holding.

A clear statement of the hypothesis being tested.

Description of dependent variables and their indicators.

Essential description of the experimental group.

Characteristics of the control group and the principles of its selection

Description of the experimental situation.

Characteristics of the experimental conditions.

The course of the experiment, i.e. his setting:

A) before the introduction of the experimental factor;

B) in the process of entering it;

C) after its introduction;

D) after the end of the experiment.

Evaluation of the purity of the experiment and the instruments used.

Conclusion on the reliability of the hypothesis.

Other conclusions.

Data on the drafters of the protocol and the degree of their agreement.

Date of signing the protocol.

Since the experimental method is more complex than others, mistakes are often made in its application. Let's name some of the most common:

The experiment is carried out to obtain information that can be obtained in other, simpler ways.

An experiment is assigned an included or standardized non-included observation.

No organic connection conducted experiment with the purpose, objectives and hypotheses of the study.

Ambiguity or other significant significant inaccuracy in the formulation of the hypothesis submitted for experimental verification was admitted.

The theoretical system of variables is built incorrectly, causes and effects are confused.

The experimental factor (independent variable) is chosen arbitrarily, without taking into account the fact that it should play the role of a determinant and be controllable by the researcher.

The independent and dependent variables were not adequately reflected in the empirical indicators.

The effect on the dependent variables of factors not included in the independent variable is underestimated.

The experimental situation is not clearly defined, due to which the experiment is carried out in violation of its conditions.

Subjective assessments of the experimental situation prevail over objective characteristics.

The experiment revealed the following important properties experimental group that were not known prior to its initiation.

The control group does not represent an analogue of the experimental group in terms of parameters essential for the study.

Control over the course of the experiment was weakened and / or ineffective.

The experimental instrumentation is aimed only at fixing certain data (like an observation instrument), and not at keeping the experiment clean.

The conclusions of the experimenters are adjusted (fitted) to the hypothesis without sufficient grounds.

social experiment sport

References

(1) Ryvkina R.V., Vinokur A.V. Social experiment. Novosibirsk, 1966;

2.Kupriyan A.P. Methodological problems of social experiment. M., 1971;

A.P. Kupriyan The problem of experiment in the system of social practice. M., 1981;

4. Working book of a sociologist. M., 1983, Yadov V.A. Sociological Research: Methodology , program, methods. M., 1987;

Vihalemm P.A. Experiment in sociological research // Methods of collecting information in sociological research. Book 2. M., 1990;

Andreenkov V.G. Methods of data collection // Sociology. The basics general theory(under the editorship of Osipov G.V., Moskvicheva L.N.). M., 1996. P.A. Vihalemm.

Tutoring

Need help exploring a topic?

Our experts will advise or provide tutoring services on topics of interest to you.
Send a request with the indication of the topic right now to find out about the possibility of obtaining a consultation.

On the possibility of using an experiment in social sciences, he expressed in his book "The Philosophical Experience of Probability", where he called for the use of methods based on observation and calculation in political and moral sciences, which already at that time proved to be good in the natural sciences. The essence of Laplace's idea, which was later taken up by many researchers, was to use such methods of a probable approach in the study of society as sampling, creating parallel control groups, etc. He argued that it was experiment and the theory of probability that could raise the status of the social sciences, bring them to the ideals of natural science.

However, it was only at the beginning of the 20th century that the field of application of experiments in the social sciences began to gradually expand. This was primarily associated with the rapid growth of empirical social research, with the formation and development of survey procedures, observation, text analysis, with the development of mathematical logic, statistics and probability theory, when means and methods of obtaining quantitatively defined information began to appear in the research arsenal. ...

Experiment- this is a research method with the help of which, by means of systematic control of the conditions of the studied situation, hypotheses about the causal relationships of phenomena are checked and tested. It is based on the development of a certain hypothetical model of the phenomenon under consideration, in which the main parameters, their interrelationships and connections with other phenomena, as well as the cause-and-effect direction of the relationships are highlighted. On the basis of this model studied in the experiment, the phenomenon is described as a system of variables, among which an independent variable (experimental factor) stands out, the action of which is subject to the control and control of the experimenter and which acts as a hypothetical cause of certain changes in the dependent variable. This general definition of experiment extends to its characterization in the field of sociology.

Therefore, the main features of a social experiment are:

Active intervention of the researcher (modeling) in the system of the studied phenomena;

Systematic introduction of a relatively isolated experimental factor, its variation, possible combination with other factors;

Systematic control over all significant determining factors;

The effects of changing the dependent variables should be measured and unambiguously reduced to the influence of the independent variables (experimental factor).

It is difficult for sociologists, of course, to fully fulfill all the requirements for an experiment that have formed in the natural sciences, but this must be striven for in order to be sure that it is the factor studied in this experiment that is the cause of the registered change (the so-called internal validity). and the revealed dependence is natural under certain conditions, and the conclusions obtained can be extended to non-experimental situations (i.e., external validity).

The possibility of using an experiment in the study of social phenomena depends on the specifics of the object of research and on how much the nature of the reaction of a social group to planned changes has been studied. Therefore, the task of a sociological experiment is to measure indications for certain factors in artificially created and controlled conditions.

At the same time, one cannot ignore those general requirements for experiment that have been developed in the process of sociological practice. The use of a sociological experiment should contribute to the growth of knowledge, the planned conditions of the experiment, its structure should not violate the fundamental principles of the object under study, the experiment should be controllable. These requirements can be met if a well-prepared experimental design is in place.

The plan of the experiment is developed depending on the specifics of the tasks and hypotheses of the research, on the one hand, and the material, social and temporal possibilities of conducting the experiment, on the other. The development of research tasks and hypotheses begins with the formulation of the goal of the experiment and the formulation of the main hypothesis, which is due to the relevance of certain social problems and the degree of their scientific elaboration.

After the formulation of the main hypothesis, it is concretized and operationalized: the goals of the experiment are determined, specific tasks are put forward, and working hypotheses are formed. For their verification, experimental and non-experimental possibilities of fixing their values are distinguished.

The structure of a social experiment can be represented as follows:

1. Experimenter. This is usually the researcher or group of researchers who design and conduct the experiment.

2. Experimental factor (or independent variable) - conditions or a system of conditions that are introduced by the sociologist. The independent variable will be controllable if its direction and intensity of action are in accordance with the program settings.

Such a variable will be controllable if its qualitative and quantitative characteristics are revealed within the experimental program.

3. Experimental situation - a situation that is created in accordance with the research program for the experiment. Under experimental conditions
the situation does not include the experimental factor.

4. An experimental object is a group of individuals who have agreed to participate in an experimental study.

The structure of the experiment is directly related to the stages of its organization and implementation.

The first stage is theoretical. In the course of it, the sociologist forms the research problem, defines the object and subject, experimental tasks and hypotheses of the research, the problem is formed on the basis of the main research task. The objects of research are representatives of various social groups, collectives or communities. When formulating the subject of research, it is necessary to project an ideal prototype of the experimental situation in symbols (signs).

At the second stage, the development of an experiment program is carried out. An experiment program is a document in which all its procedures and elements are developed.

The third stage is the implementation of the experiment program. In the course of it, the direct implementation of the intended experiment, its stages and all procedures is carried out, i.e. the implementation of the experimental situation. ...

At the fourth stage, the analysis and evaluation of the results obtained during the experiment is carried out. As a result of such an analysis, it is determined whether the goal has been achieved, whether the research hypothesis has been confirmed. In addition to the main results, it is possible to obtain related results that can be used in further scientific research.

The methodology distinguishes the following types of social experiments. By the nature of the object and subject of research - sociological, economic (economic), legal, socio-psychological, psychological, aesthetic experiments.

The differences between these types of experiments are determined by the specifics of the respective scientific disciplines.

By the nature of the research procedures, there are real and mental experiments. Real - when hypotheses are tested by systematic management of the conditions of social activity. In contrast, in a thought experiment, hypotheses are tested not by real phenomena, but by information about them. Very common types of social experiment are, for example, both real and mental economic (economic) experiments, which according to the specifics of the task (dominant function) are divided into scientific and applied, according to the nature of the experimental situation - into more or less controlled, etc. ...

By the nature and subject of research, among thought experiments, one can distinguish an experiment based on a sign model, other types of experimental (de-termination, causal) analysis of previously collected information, and ex-post-facto experiments. When conducting a model experiment, hypotheses are tested by purposeful activity not on real objects, and on their substitutes - on full-scale, artificial or iconic models (analogs).

Almost all sociological experiments are produced not on the general population of research objects, but on a natural model-sample based on the characteristics that are most significant for this research. The representativeness of the sample of subjects is at the same time a prerequisite for the representativeness of the results of the experiment.

Mathematical, statistical, cybernetic models act as symbolic models in thought experiments. Their distinctive feature is multifactoriality - a researcher, using modern computer programs, can simultaneously "play" the values of a whole complex of experimental factors. This largely promotes the solution of the problems of a comprehensive study of complex social processes, makes it possible to move from the level of description to the level of explanation and further to the level of forecasting.

A special kind of thought experiment is the so-called ex-post facto experiment. This type of research involves collecting data to test hypotheses about causation. The alleged causal relationship has already been carried out, and therefore the research is aimed at collecting and analyzing data about the events that happened, about the conditions and the alleged reasons for their occurrence.

A classic experiment of this type is the study of the American sociologist E. Christiansen, conducted in the mid-1930s, aimed at testing the following hypothesis: the higher the level of education (the number of completed classes), the more successful a person is included in economic activity(the empirical indicator of this dependent variable was the increase or maintenance of the level of wages when changing jobs). Of the 2,127 people who left St. Paul's secondary schools in 1926 (of which they graduated from high school, 997 did not graduate), in 1935 1,194 were tracked down and interviewed (671 with a school leaving certificate and 323 without it). Through the analysis of the documents, a wide range of information was obtained about the stay of these people at school.

The results of the study confirmed the hypothesis: as the number of completed grades increased, the number of people who turned out to be economically more adapted after 9 years grew - among those who graduated from 7 grades there were 74%, 8 grades - 85%, 9 grades - 90%. In order to more rigorously identify the role of school training in economic success, two maximally identical according to the data of the period were formed. schooling groups (23 people in each), differing in only one feature (independent variable) - the presence of a maturity certificate (people in the experimental group had them, people of the second group did not). Analysis of the interview data showed that in the experimental group for 9 years, among which there were years of severe economic crisis, 92% increased or retained their salaries, in the control group - 58%, which was sufficient evidence of the initial hypothesis.

Famous experts in the field of social experiment F. Chepin and E. Greenwood attached great importance to the study of E. Christiansen and the "ex-post-facto" experiment in general, predicting its wide distribution and seeing in it the possibility of social sciences to achieve accuracy equal to that of natural sciences in proving explanatory hypotheses. However, the ex post facto experiment was not widely adopted. Although, in its design, it is close to longitudinal research. "

According to the specifics of the task, they differ scientific and applied experiments. In the course of a scientific experiment, a hypothesis is tested, containing new information of a scientific nature, which has not yet found its sufficient confirmation or has not yet been proven at all. Applied experiment includes numerous processes of experimentation in the field social relations and is aimed at obtaining a practical effect.

Also, according to the specifics of the task at hand, one can distinguish projective and retrospective experiments. Projective ones are directed to the future: the researcher projects the manifestations of the alleged consequences, putting into action hypothetical causes. As a special kind of projective social experiments, one can single out a creative experiment aimed at testing a social project, fundamentally new management decisions. A real experiment is always projective, while a mental one is usually retrospective, directed into the past: the researcher manipulates information about the events that have taken place, tries to test hypotheses about the causes that caused the consequences.

According to the specifics of the task, they also differ one- and multivariate experiments. In a one-factor experiment, a hypothesis about the consequences of the influence of one independent variable is manifested, in a multifactorial experiment, a whole complex of variables in their interaction.

By the nature of the logical structure of the proof of the hypothesis, experiments are subdivided into parallel and consistent.

A parallel experiment is understood as a study in which there is both an experimental and a control group, and in which the proof of a hypothesis is based on a comparison of the state of two objects of observation (experimental and control) at the same time.

In a sequential experiment, there is no control group. The same group acts in it as a control before the introduction of the independent variable and as an experimental one - after the independent variable had (or could have had) the intended effect. The proof of the hypothesis in this case is based on the comparison of two states of the object of observation at different times.

According to the nature of the experimental situation, experiments in sociology are divided into field and laboratory. For field research, the most natural situation is characteristic (working environment, student group, class etc.), the objects investigated with its help, as a rule, retain their usual connections.

Among field experiments, controlled and natural are distinguished. A controlled experiment is understood as an experiment when a researcher has a description of the factors that make up a social object and the conditions for its existence, and enters into action an independent variable as a hypothetical cause of the changes expected in the future. V natural experiment the researcher does not pre-select or prepare an independent variable, nor does it affect the experimental group. Thus, the researcher does not actively intervene in the usual course of events, but only observes and waits until the clearly expressed change of interest to him, which will play the role of an independent variable, independently occurs in the process under study.

A laboratory experiment is a study in some kind of artificial environment. The artificiality of the latter lies in the fact that the object of observation from its usual, natural environment is transferred to an environment that allows a high degree of accuracy in observing its behavior to be achieved. A special place in conducting a laboratory experiment is occupied by the organization of the experimental situation and its management. The latter is natural and playful.

Before proceeding to the analysis of the results of the experiment, it is necessary to organize the obtained materials, i.e. determine the classification of the results obtained, their processing, which would make them accessible. The purity of the experiment, combined with the adequate use of methods of sociological analysis of the results, can guarantee that the relationship of the studied variables revealed during the experiment will not be accidental. Further, it seems important to determine how general the relationships and mechanisms identified in the experiment have. This is especially important when studying the social problems of interaction in a society of a transitional type, when making managerial decisions.

The experimental objects of observation under study will always be only a part of all possible options... If, in the course of the study, their main characteristics are changed, it is possible to establish the boundaries of the situation for which the experimentally obtained conclusions will be adequate.

The application of an experiment in sociology is associated with a number of difficulties that do not allow achieving the purity of a natural science experiment, since it is impossible to eliminate the influence of relations existing outside the investigated, it is impossible to control factors to the extent that it is possible in a natural science experiment, to repeat it in the same form. progress and results. At the same time, it is impossible not to take into account certain ethical problems, which requires increased responsibility from the researcher, since the experiment directly affects a specific person.

One of the most effective and at the same time the most laborious methods of collecting sociological information is experiment. It allows you to get very unique information, which cannot be obtained by other methods.

As a kind of in-depth, analytical sociological research and at the same time a method of collecting information about the factors affecting the state of certain social phenomena and processes, as well as the degree and results of this impact, the experiment is of great scientific and practical value.

Experiment- is “a method of obtaining information about the studied social object by influencing it by some controlled and controlled factors (variables). Requires the advancement of a hypothesis about the relationship of its various characteristics "The purpose of any experiment is to test hypotheses about a causal relationship between phenomena: a researcher creates or searches for a certain situation, activates a hypothetical cause and observes changes in the natural course of events, fixes their correspondence or inconsistency with assumptions, hypotheses ... The experiment is also the most reliable means of solving many practical problems associated with assessing the effectiveness of social and political programs.

The experimental proof of the hypothesis is based on logical circuits developed by J.S. Mill. In his book "System of Logic", first published in 1843, Mill formulated 4 methods of proving causal relationships: the method of a single difference, concomitant changes, a single similarity, and residuals.

In the social sciences, the method of single difference is most often used, which consists in comparing two complex phenomena (systems, processes), differing only in that one of them has a hypothetical cause, and the other does not.

A modification of this system of proof is the scheme of accompanying changes, which differs from the above only in that the action of the independent variable is performed many times and with different intensities, which should lead to corresponding changes in the dependent variable.

The logic of the scheme of the only similarity is different - the two compared systems differ in all parameters, except for the experimental variables. In the study of social phenomena, this scheme is almost never used, because it is even more difficult to prove that two systems differ in everything except the action of the experimental factor than to prove that they are similar in everything except this.

The method of residuals, which is based on already proven causal relationships between variables, is also inapplicable in social research.

An interesting fact is that JS Mill himself, and behind him O. Comte, E. Durkheim, M. Weber and others denied the possibility of using the experimental method in the study of social phenomena. The problem is that social phenomena are too complex and changeable for this, it is impossible to clearly identify the effects of one specific factor, as well as an unambiguous interpretation of the behavior of a person or a social community.

Complexity, multifactorial and multi-layered social processes; difficulty, often impossibility of their formalization and quantitative description; holistic, systemic nature of dependencies; mediation external influences through the human psyche, including through a predisposition to a certain attitude towards behavior - all this makes it difficult to empirically identify and prove causal dependencies. A number of social processes have been little studied for the advancement of explanatory hypotheses. And if they exist, a real social experiment is often impossible for political, economic, ethical and other reasons. Most important social situations cannot be created at the request of the researcher, just as in already existing situations, causal connections cannot be experimentally verified, the implementation of which could lead to undesirable consequences. Many problems arise in connection with the representativeness of experiments carried out with a certain specific group of people under certain specific conditions. It is very difficult to establish the boundaries of the possible extension of the experimental results to other groups and conditions, especially when the subjects know that the experiment is underway and their reactions are recorded. Organizational problems associated with conducting an experiment are quite complex and troublesome. "

Experiment in the form in which it is applied in the natural sciences cannot be applied in the social sciences. In sociology, the general logic of the experiment is that, by choosing an experimental group (or groups) and placing it in an unusual experimental situation (under the influence of a certain factor), to trace the direction, magnitude and stability of changes in characteristics called control. In this sense, the experiment is something like closed system, the elements of which begin to interact according to the "script" written by the researcher.

There are three requirements to follow when designing an experiment. First, the characteristics that are most important with

point of view of the problem under study. Secondly, the change in the control characteristics should depend on those characteristics of the experimental group (or the environment of the experiment) that are introduced or changed by the researcher himself. Such characteristics are called factorial. The characteristics "not participating" in the experiment are called neutral. Their "fate" in the course of the experiment can develop in different ways. In some cases, they can change without "outside" influence (then they get the name variables), in others - to remain unchanged (then they are called constant). For the final results of the experiment, both are not of great importance. The main thing is that changes in the neutral characteristics of ns are reflected in the control ones. And thirdly, the course of the experiment should not be influenced by those phenomena that do not relate to the experimental situation, but are potentially capable of changing it.

The listed conditions actualize the problem of choosing an experimental group.

Experts identify several types of experiments and methods for selecting experimental groups.

By the nature of the object and the subject of research, sociological, economic (economic), legal, socio-psychological, pedagogical, psychological experiments are distinguished. In this section, we are primarily interested in sociological experiments.

According to the logical structure of proving hypotheses, a linear and a parallel experiment are distinguished.

In a linear experiment, the same group is analyzed, which is and control(its original state), and experimental(its state after changing one or more characteristics). That is, even before the start of the experiment, the control, factorial and neutral characteristics of the object are clearly recorded. After that, the factorial characteristics of the group (or the conditions of functioning cc) change, and after a certain predetermined period of time, the state of the object is again measured according to its control characteristics.

It is very important that in the process of a linear experiment the influence of interfering factors on the object of analysis is excluded.

In a parallel experiment, two groups are simultaneously involved: control and experimental. Their composition should be identical for all controls, as well as for neutral characteristics that can affect the outcome of the experiment (first of all, these are socio-demographic characteristics). The characteristics of the control group remain constant throughout the entire period of the experiment, while the characteristics of the experimental group change. Based on the results of the experiment, the control characteristics of the two groups are compared and a conclusion is made about the causes and magnitude of the changes that have occurred.

For a successful experiment, the correct selection of its participants plays an important role. Three methods are used in applied sociology: pairwise selection, structural identification, and random selection.

All three methods are presented with one general requirement: they must guarantee the legitimacy of the distribution of the results of the experiment to the object, the change in the characteristics (or operating conditions) of which is envisaged in the future. This object acts as a general population, from which the experimental group is selected. In addition, the use of the appropriate method for selecting the group (s) is to some extent predetermined by the model of the planned experiment.

Pairwise selection method. It is used primarily in a parallel experiment. Its essence is as follows. Two groups are selected from the general population in such a way that they are identical in terms of neutral and control, but differ in factor characteristics. For both groups, the same conditions are created, and after a while the effect of the experiment is measured by fixing and comparing the parameters of control signs in both groups.

The method of structural identification can be used in both linear and parallel experiments. In this case, in a linear experiment, the group is selected so that it is a micromodel of the general population in terms of neutral and control characteristics. Such selection can be carried out on a quota sampling basis. In turn, in a parallel experiment, the structures of the experimental and control groups are aligned according to the same characteristics. For example, the size of the two groups is 50 and 90, respectively. It is known that in the first experimental group 70% of the group members (35 people) have secondary education, and 30% (15 people) have higher education. In the second (control) group, their share is equal - 50% each (45 people each). Suppose that in order to create an experimental situation, we need to bring the structure of the control group in accordance with the structure of the experimental one according to the characteristic “level of education”. Making simple arithmetic calculations, we find that the control group should include 60 people: 42 (70%) with secondary education and 18 (30%) with higher education.

The random sampling method is identical to the previously discussed random sampling methods with a predetermined size. As a rule, it is used in field experiments with a large (up to several hundred) size of the experimental group.

According to the method of carrying out, real and mental experiments are distinguished. In a real experiment, explanatory hypotheses are tested by systematic management of the conditions of social activity. In thought experiments, hypotheses are tested not by real phenomena, but by information about them.

According to the specifics of the task at hand, theoretical and applied experiments are distinguished. The first are aimed at acquiring new scientific knowledge, information of a methodological nature, and the second - to obtain a practical effect.

Depending on the orientation towards the past or the future, projective and retrospective experiments are distinguished. Projective experiments are directed towards the future: the researcher projects the manifestations of the alleged consequences, putting into action hypothetical causes. As a special kind of projective social experiments, one can single out a creative experiment aimed at testing a social project, fundamentally new management decisions. A real experiment is always projective, while a mental one is usually retrospective, directed to the past: the researcher manipulates information about the events that have taken place, tries to test hypotheses about the causes that caused the consequences. Thought experiments can also be projective if they are used to predict future events and if their conclusions are verified by real experiments and other research methods.

According to the nature of the experimental situation, experiments are divided into field and laboratory. In a field experiment, an object (group) is in the natural conditions of its functioning (for example, a production group). In this case, the members of the group may or may not be aware that they are participating in the experiment. The appropriate decision in each case depends on how much awareness can influence the course of the experiment.

In a laboratory experiment, the situation, and often the groups themselves, are formed artificially. Therefore, team members are generally aware of the experiment.

In both field and laboratory experiments, interviewing and observation can be successfully used as additional methods of collecting information. Their results provide a basis for the researcher to decide whether to intervene in the course of the experiment or observe it before complete completion without interference.

The preparation and conduct of the experiment involves consistent solution a number of questions:

1) determination of the purpose of the experiment;
2) selection of an object (objects) used as an experimental, as well as a control group (groups);
3) highlighting the subject of the experiment;
4) the choice of control, factorial and neutral signs;
5) determination of experimental conditions and creation of an experimental situation;
6) formulation of hypotheses and definition of tasks;
7) the choice of indicators and a method for monitoring the course of the experiment;
8) determination of the method of fixing the results;
9) the choice of the criterion for evaluating the effectiveness of the experiment. "

On the basis of the selected variables and material, social, temporal possibilities of conducting research, an experimental plan is developed. In social research, 4 plans are most often used: 1) a "before-after" experiment without a control group; 2) a "before-after" experiment with a control group; 3) an experiment "only after" with a control group;

4) an experiment "supposedly before - after" with a control group.

An experiment "before-after" without a control group belongs, according to the logic of proving a hypothesis, to sequential experiments. The researcher creates or searches for an experimental situation before the introduction of the experimental factor, and after its impact, the characteristics of interest to the experimenter are measured - the independent variable and related factors, the characteristics of the situation. The hypothesis is tested by comparing the values of the dependent variable before and after the influence of the experimental factor.

The "before-after" plan with the control group is parallel in the logic of the hypothesis proof. It presupposes the formation or search for two groups equal in their main indicators, in one of which (experimental) the experimental factor is introduced, and in the other (congressional) it is not. In both groups, measurements of the characteristics of interest to the researcher are made before and after the action of the experimental factor. The hypothesis is considered proven if there were changes in the dependent variable in the experimental group, but in the control group, they were not observed.

In the "only after" experiment with the control group, there are no problems associated with preliminary measurements of the variables ("the effectiveness of the first measurement"). The values of the variables are fixed only after the influence of the experimental factor. The hypothesis is tested by comparing the values of the dependent variable in the experimental and control groups.

There are a number of experimental designs, where by combining measurements "before" and "after" in aligned groups, they try to reduce Negative influence preliminary measurements, on the one hand, and compensate for the lack of data on the position before the influence of the experimental factor, on the other.

Such experimental plans include a "supposedly before - after" with a control group. This plan is characterized by the fact that in both groups only one measurement is made, but at different stages of the experiment - in the experimental group - before its impact. The hypothesis is verified by comparing the results of these measurements, i.e. the experimental and control groups are considered, in essence, as one, so that the initial position can be measured in one group, and the final position in the other, and still consider the difference in these states as a result of the influence of the experimental factor.

All the experimental designs described above are one-factor (single-variant). A multifactorial (multivariate) design involves the influence of two, three or even more experimental factors. The multivariate design allows you to identify the interaction of experimental factors.

Social experiment is one of effective techniques research of control systems. Its significance lies in the fact that it helps to reveal social processes in all their complexity and diversity, gives the governing bodies reliable information, allowing them to assess the effectiveness of a particular management system. In this regard, let us mention the Hawthorne Experiments (1927-1932). Studying the influence of various factors (conditions and organization of work, wages, interpersonal relationships and leadership style) to increase labor productivity at an industrial enterprise, the organizer of the experiment E. Mayo concluded that the human factor plays a special role in production. Generalization of empirical data allowed E. Mayo to create a social philosophy of management.

Social experiments play a special role in management. The need for an experiment arises when ordinary sources of information are not enough for making management decisions, when it is difficult to foresee all parties in advance. practical application new solution. During a social experiment, the project management decision acquires concreteness and purposefulness, which makes it possible to judge the degree of its social significance.

In the process of developing a management decision, there can always be various options, from which it is necessary to choose the most effective one for a particular situation. Therefore, a social experiment acts as a means of searching for specific forms best solution assigned tasks. Experiment provides an opportunity to practically experience various options management decisions and, based on the analysis and synthesis of information about the functioning of the tested options, select the most effective of them for the correct management act.

Social experiment is scientific laboratory, which can be used for scientific forecasting of the further course and development of management decisions. With the help of a social experiment, the subject of control has the ability to solve not only urgent and current problems, but also to foresee the course of events. Therefore, experiment is a concrete form social forecasting... It often gives rise to new hypotheses that can be used for subsequent experiments. In other words, experiments of the first order can become the starting point for experiments of the second, third, etc. order, which will help to obtain more extensive information about the deep processes of public life and contribute to the development and adoption of optimal management decisions. "

We examined various methods of collecting sociological information and want to draw the attention of readers to the problem of the choice of research methods.

Methods of collecting information have different cognitive capabilities, advantages and disadvantages.

First, none of the data collection methods is universal in relation to the subject of sociological research. It is the specificity of the reflection of objective reality in information sources that requires a sociologist to use complex different methods for the development of the most diverse sources of information and, ultimately, for the most complete comprehension of the essential properties of the studied subject. At the same time, when conducting a “monomethodical” study, a sociologist must observe the boundaries of interpretation of the data obtained, without exceeding the cognitive capabilities of the method used and the source of information with its inherent features of reflecting objective reality.

We encounter such a mistake in sociological reports, conclusions and recommendations when it comes to interpreting survey data that characterize the reflection of the reality being studied in the minds of the respondents as an unambiguous correspondence to this reality itself. This problem is especially aggravated in sociological research aimed at studying socially disapproved types of activity or complex social phenomena associated with everyday consciousness by complex mediated reflection mechanisms.

Secondly, the specificity of the reflection of the studied reality in information sources gives rise to many of its technical varieties within the framework of each of the main methods. Moreover, each technical variety the method is not indifferent to its cognitive capabilities, has its pros and cons that affect the quality of the information received, the economic and organizational costs of the study. "

On the one hand, a priori (initial) knowledge of the sociologist about the cognitive capabilities of each of the methods of data collection, on the other hand, a priori knowledge about the subject of research, the features of a particular research situation, act as the basis for determining the methodological research strategy. The transition from theoretical premises of research to the stage of collecting empirical information is implemented in methodological solutions at various levels.

1. At the level of the methodological research strategy as a whole, decisions are made about the necessary and sufficient number of methods, the cognitive capabilities of which in the aggregate are assessed as adequate to the objectives of the research and the specifics of the research situation.
2. At the level of a separate method, decisions are made on the use of such technical and organizational varieties of the method, the cognitive capabilities of which are assessed as adequate to the particular cognitive tasks solved using this method.
3. At the level of a separate research tool (questionnaire, plan, interview, instruction or interviewer's report, block of questions, separate question, code card, etc.), decisions are made about the adequacy of the cognitive tasks of particular procedures for identifying, changing, registering fragments of the reality being studied.

Thus, data collection methods are not just a conglomerate of tools that can be arbitrarily (or not used) by the researcher, depending on organizational resources and personal preference. The choice of data collection methods is dictated by the objective nature of the studied social phenomena, the specifics of the reflection of their properties in potential sources of the required information. At the same time, the choice of data collection methods quite rigidly sets the boundaries of meaningful interpretation of the received empirical information. "

Gorshkov M. K. Sheregi F. E. Applied Sociology. M., 2003.S. 152.

Yadov V.A.Strategy of sociological research. M, 1999 S. 358-361.

Social management: Textbook / Ed. D.V. Valovoy. Moscow: ZAO Business School Intel-Sintez, Academy of Labor and Social Relations, 2000. P. 183-184.

Methods of collecting information in sociological research. Book. 1. / Resp. ed. V.G. And-reenkov, O. M. Maslova. Moscow, 1990, p. 41-42.

An experiment in sociology. Social experiment in sociology

Features of social experiments

Theory and methodology sociological research

Tutoring

The last

You need to know