Crenças e atitudes na física e sua aprendizagem: análises estatísticas e teóricas a partir do questionário CLASS

Abstract

The Colorado Learning Attitudes about Science Survey (CLASS) is a questionnaire aimed at assessing students' beliefs and attitudes towards physics and the learning of physics. In this study, we analyze CLASS from two perspectives: its theoretical-methodological aspect as a test of epistemic beliefs and the quantitative analyses that can be conducted through its application. By reviewing the concepts of beliefs and attitudes, we observed several uncritical uses of these terms in works that utilized CLASS, which is an expected result due to the lack of a theoretical framework in the discussions surrounding the questionnaire. Analyzing CLASS considering research on epistemic development, we discuss the problems caused by the absence of a theoretical basis, which complicate analyses and comparisons, as well as increase confusion around already complex terms such as epistemic beliefs. Considering studies from the past two decades in the field of epistemic development, we suggest that CLASS be viewed as a questionnaire that evaluates aspects of epistemic cognition. We analyzed CLASS data from 465 students, observing differences based on course and gender. Significantly higher scores were obtained in the physics course compared to engineering and other exact science courses, and the same was observed regarding gender, with women scoring significantly lower than men. We found small but significant effects of a relationship between the CLASS score and the score on a conceptual electromagnetism test. Next, we present a descriptive analysis of the distribution of CLASS responses from physics and engineering students. Finally, we employed three main tools and statistical methods in the study: factor analysis, network analysis, latent profile analysis, and multiple imputation. In the factor analysis and network analysis, we grouped questions into factors or clusters (in the case of networks), which indicate similarities between the questions. In the latent profile analysis, we grouped individuals into profiles that indicate similarities in their responses.