A Method Detecting Student’s Flow Construct during School Tests through Electroencephalograms (EEGs): Factors of Cognitive Load, Self-Efficacy, Difficulty, and Performance

Abstract

This study gathers and examines information about the flow state’s emergence during tests and its factors using an electroencephalogram (EEG) to establish a method and reveal an individual student’s flow construct. Through a single-case experimental design and 766 test items, multiple measurements were performed on a 14-year-old junior high school science-gifted student. During the test, self-efficacy, item difficulty, cognitive load, and test performance (long-term test performance [LT-tp] and short-term test performance [ST-tp]) were examined to establish the construct of EEG-detected, real-time flow states (EEG-Fs). Based on the chi-square test of independence results, the EEG-F had a significant correlation with the student’s cognitive load, self-efficacy, LT-tp, and item difficulty. Furthermore, a J48 decision tree analysis and logistic regression revealed four inhibiting and two inducing conditions affecting the emergence of EEG-Fs. The two inducing conditions included (1) high self-efficacy with a low cognitive load (odds ratio (OR) = 3.7) and (2) high cognitive load when combined with high self-efficacy and LT-tp for low-difficulty items (OR = 3.5). The established method and findings may help teaching designers or automated teaching applications detect the individual student’s flow construct to select appropriate test tasks accordingly, resulting in an optimal experience and better achievements.

Novel Feature-Based Difficulty Prediction Method for Mathematics Items Using XGBoost-Based SHAP Model

“…Using the known difficulty information of the exam items as labels, a machine learning model is constructed to predict item difficulty. Traditional machine learning models include regression analysis [3], support vector machines (SVMs) [34], decision trees [35,36], Random Forests [37], and shallow BP neural networks [38]. On the other hand, deep learning models such as Convolutional Neural Networks (CNNs) [39], Recurrent Neural Networks (RNNs) [40] and Long Short-Term Memory (LSTM) [41] neural networks are also used.…”

A Preliminary Study of the Efficacy of Using a Wrist-Worn Multiparameter Sensor for the Prediction of Cognitive Flow States in University-Level Students

“…Flow is a construct of interest across many different contexts and fields, such as serious gaming [28], collaboration [29], exercise [30], education [31][32][33], and even work [34], to name a few. While flow is typically measured through subjective means, there is no widely agreed upon "best" measure [35].…”