Predicting attribution of letter writing performance in secondary school: A machine learning approach

Abstract

The learning research literature has identified the complex and multidimensional nature of learning tasks, involving not only (meta) cognitive processes but also affective, linguistic, and behavioral contextualized aspects. The present study aims to analyze the interactions among activated domain-specific information, context-sensitive appraisals, and emotions, and their impact on task engagement as well as task satisfaction and attribution of the perceived learning outcome, using a machine learning approach. Data was collected from 1130 vocational high-school students of both genders, between 15 and 20 years of age. Prospective questionnaires were used to collect information about the students’ home environment and domain-specific variables. Motivation processes activated during the learning episode were measured with Boekaerts’ on-line motivation questionnaire. The traces that students left behind were also inspected (e.g., time spent, use of provided tools, content, and technical aspects of writing). Artificial neural networks (ANN) were used to provide information on the multiple interactions between the measured domain-specific variables, situation-specific appraisals and emotions, trace data, and background variables. ANN could identify with high precision students who used a writing skill, affect, and self-regulation strategies attribution on the basis of domain variables, appraisals, emotions, and performance indicators. ANN detected important differences in the factors that seem to underlie the students’ causal attributions.