Can the way a learner interacts with an exercise help reveal when their confidence in completing a task is low? An exploratory study investigates that question in technology-enhanced learning environments, using interaction traces alongside self-reported self-efficacy.
What the researchers collect
The authors conduct an experiment across two universities and collect both activity traces and learners' own reports. They develop a machine-learning pipeline with multiple classification models to investigate features associated with low self-efficacy.
The aim is to help identify learners who may benefit from support early in the learning process. That goal should be distinguished from an established intervention: detecting a pattern in a study does not show that automatically acting on it improves a student's outcomes.
Why timing matters
The paper emphasizes that self-efficacy can change during learning. A label collected at one moment may therefore describe a different state from the one reflected in later interactions. The authors highlight the importance of closely monitoring these beliefs for prediction performance.
For a learning-analytics team, that suggests a concrete evaluation question: how far apart are the activity being analyzed and the self-report used as its label? A model can appear to predict a psychological state while instead learning features of a task, session or measurement schedule.
What the early evidence can support
The researchers describe the work as exploratory and present initial findings and future directions. The captured evidence does not justify a universal accuracy claim, a diagnosis of an individual student, or deployment as an automated gate for educational opportunities.
A useful pilot would compare the model's signals with appropriately timed student feedback and educator judgment. Review errors across learners and exercises, and give people a way to correct an interpretation. Interaction traces can contribute evidence, but the student's own account and the teaching context remain essential when deciding what support would help.
Comments