Study title

Personal qualities and skills such as perseverance and deep learning are important to assess and monitor for educational purposes, but their measurement in cross-cultural assessments is challenging. Conventional self-reported measures suffer from biases which can produce paradoxical results and pose difficulties to draw meaningful comparative inferences. We propose a new source of behavioral measures constructed unobtrusively from computer-generated logs in addition to self-reported measures. Using data from two large-scale educational surveys, we tested the convergent and predictive validity of self-reported and behavioral measures, and found that the behavioral measures are resistant to the self-presentation style or reference group effect, hence can compensate for the limitations of self-reports in cross-cultural analysis. We also illustrate the complementary nature of both measures, and encourage researchers to apply them integratively in educational evaluations.