心理学における情報の概念について
スポンサーリンク
概要
- 論文の詳細を見る
The statistical treatment of psychological data has been considered so important as it is an essential part of experimental methodology. But the measuring procedures of responses do not generally correspond to the basic structure of statistics; for example, there exists neither the independence among measured values nor the normality of the distribution of judgment errors nor what not. In order to avoid this inconsistency, various efforts have been paid. One of them is the establishment of the concept of Information in psychology. The concept of the transmitted information in the judgment of a stimulus pattern has recently become more important in the treatment of perceptual data. It is free from the statistical independence among measured values, and plays rather an essential role in the estimation of a psychological intervening variable, judgment. Since the judgment is considered as a system of communication between a stimulus pattern and respose categories, an information transmitted from stimulus to response is to represent the uncertainty or the freedom of choice with respect to the judgment. If the judgment has only one possible choice, the amount of the transmitted information will be zero, and the judgment is, therefore, simply the function between stimulus and response. On the contrary, in case of the judgment having multiple possible choices, an intuitive probability distribution pattern can be contracted for every choice. Such a construction is a theoretical model of judgment. If the model of the judgment is limited to the linear combination of all informations which affect the categorization of responses, it will mathematically be expressed In other words, informations of responses observed by an experimenter are broken up into the linear combination of all informations. Let x be a series of stimuli, and y be response categories, then the response information H(y) will be as follows: H(y)=H_x(y)+T(x;y)+A(xy) where A is interaction information, T is the amount of information transmitted from stimulus to response (freedom of choice in respect to pure judgment), and H_x(y) means the error term of response information. In Shannon's theory (19), H_x(y) is the white noise, whereas in psychology, as McGill has pointed out, it is the more systematic bias of information source, that is, another source of information of judgment. In Garner and Hake's paper (6), it was already found that there was response equivocation, E_r, in judgment. McGill has developed this idea, analysing it into two parts, i.e. interaction information and error terms (16). And then, he has constructed the concept of information population, distingushed from the sample information, and has also developed the method of estimating an information population, testing the hypothesis of the population with regard to the multivariate information sources. When the model of judgment is considered on the basis of these theories of information, it can be said that psychophysical measurements are the double estimation of a stimulus pattern made by an experimenter. The judgment receives informations from the stimulus and transmits them to the response. The experimenter, however, observes only the transmitted informations, he has to guess the structure of the observer's judgment. An adequate model of judgment may be fruitful for such guessing. As to the perceptual data, such an assumption as independence of responses or normality of the distribution of measured values, may hardly be made for the practical measurement, but the informational analysis of them does not require such assumptions. This will, therefore, be a step towards the advance of the statistical treatment of psychophysical data.
- 東京女子大学の論文
- 1955-12-20