Lecture
Natural, artificial, and laboratory experiments.
This classification criterion is set by the differences among independent variables in terms of their correspondence either to the conditions of real human life activity or to theoretical concepts operationalized at the level of specific methodological means.
Artificial experiments: the experimental situation models a real one but allows the relationship between dependent and independent variables to be controlled better by eliminating or stabilizing a number of side variables that are usually present under real conditions of life activity.
Laboratory experiments, like artificial ones, are conducted under specially created conditions. They presuppose purifying the conditions of the experiment so that single independent variables can be varied.
The difference between artificial and laboratory experiments lies in what the experimental model represents: a real situation or a theoretical model. This difference can be traced in the forms in which artificial and laboratory experiments are conducted: in the former case, games and simulators, where the purification of conditions must not affect the quality of the process under study.
It is precisely artificial experiments that Gottsdanker called «world-improving».
They are conducted when the mere reproduction of a real situation does not make it possible to render the experiment internally valid. However, a question arises: can the results of such an experiment be applied to reality? Thus, the problem of external validity comes to the fore. In artificial experiments it is possible to raise internal validity. Three ways of improving the real world are described that make this possible: eliminating systematic confounding; the possibility of obtaining the required amount of data in a shorter time and thereby increasing the reliability of the experiment; and the possibility of reducing non-systematic variability of the data and, consequently, their scatter, which also ensures higher reliability.
Achieving correspondence of the independent variable in an experiment is, as a rule, fairly simple. The conditions introduced must be either typical of real situations or quite probable.
The correspondence of the dependent variable was assessed on the following three points. (1) Does the subject's work in the experiment correspond to their real-life activity? (2) Do the measured indicators reflect the most important aspects of this activity? (3) Is the method of presenting the measurement results adequate?
In artificial experiments, questions also arise about the correspondence of additional variables that are stable in their level. In a whole range of cases such variables are key, and their level must correspond to the real world. One should also try to reproduce in the experiment those additional actions (relative to the main task) that in reality are performed simultaneously with it. In addition, one should determine how the absence of emotional tension (which, as a rule, is characteristic of the corresponding real situations) will affect the external validity of an artificial experiment. The consequences of presenting the subject with all the experimental trials over a short period of time (compressed compared with an experiment that duplicates reality) should be specifically analyzed.
Field experiments (examples)
- the trial of a new physics teaching program in universities – the Harvard Project Physics. Randomization was implemented as a strategy of selection from the entire population of those studying the subject (teachers and students). The subject knowledge of the students in both groups turned out to be the same, but in terms of indicators of emotional attitude the new program received a higher rating.
Comments