Affective States and State Tests: Investigating How Affect and Engagement during the School Year Predict End-of-Year Learning Outcomes

Type: Evidence | Proposition: A: Learning | Polarity: | Sector: | Country:

Affect and behaviour detectors were used to estimate student affective states and behaviour based on analysis of tutor log data. For every student action, the detectors estimated the probability that the student was in a state of boredom, engaged concentration, confusion, or frustration. They also estimated the probability that the student was engaging in off‐task or gaming behaviours.

  • Boredom during problem solving was negatively correlated with performance
  • Boredom during scaffolded tutoring was positively correlated with performance
  • Confusion during problem solving was negatively correlated with performance
  • Confusion during scaffolded tutoring was positively correlated with performance
  • Engaged concentration was associated with positive learning outcomes
  • Frustration was associated with positive learning outcomes
  • Gaming the system was associated with poorer learning,

A unified model was used to predict student standardized examination scores from a combination of student affect, disengaged behaviour, and performance within the learning system. This model achieved high overall correlation with standardized exam scores.

Building on this work could produce positive evidence for ‘Proposition A: learning analytics improve learning outcomes’ and ‘Proposition B: Learning analytics improve learning support and teaching’.

Citation: Pardos, Zachary A, Baker, Ryan S, & Gowda, Sujith M. (2014). Affective states and state tests: investigating how affect and engagement during the school year predict end-of-year learning outcomes. Journal of Learning Analytics, 1(1), 107-128. | Url: http://epress.lib.uts.edu.au/journals/index.php/JLA/issue/view/307