Emergent behaviors in computer-based learning environments: Computational signals of catching up

Abstract

Self-regulative behaviors are dynamic and evolve as a function of time and context. However, dynamical fluctuations in behaviors are often difficult to measure and therefore may not be fully captured by traditional measures alone. Utilizing system log data and two novel statistical methodologies, this study examined emergent patterns of controlled and regulated behaviors and assessed how variations in these patterns related to individual differences in prior literacy ability and target skill acquisition. Conditional probabilities and Entropy analyses were used to examine nuanced patterns manifested in students’ interaction choices within a computer-based learning environment. Forty high school students interacted with the game-based intelligent tutoring system iSTART-ME, for a total of 11 sessions (pretest, 8 training sessions, posttest, and a delayed retention test). Results revealed that high and low reading ability students differed in their patterns of interactions and the amount of control they exhibited within the game-based system. However, these differences converged overtime along with differences in students’ performance within iSTART-ME. The findings from this study indicate that individual differences in students’ prior reading ability relate to the emergence of controlled and regulated behaviors during learning tasks.