Design and Evaluation of a Grouping Algorithm to Reduce Bias and Increase Individual Speech in Collaborative Learning in Upper Elementary School

Abstract

we design and evaluate a group formation method that decreases the bias in the amount of individual speech and increases the amount of individual speech during collaborative learning in inquiry learning in upper elementary school. Based on the hypothesis that children's interactions increase or decrease the amount of individual speech, we created 10,000 different group combinations by inputting children's interests, abilities, and personality traits data, and calculated the predicted amount of individual speech, the average predicted amount of group speech, and the bias of group speech for each group. For each group, we calculate the predicted amount of individual speech, the average value of the predicted amount of speech within the group, and the bias of speech within the group. The results showed that there was a correlation between the predicted and measured amounts of speech and that a certain amount of speech could be predicted, and that there was an increase in the amount of speech, especially among children who usually do not speak much.