Focal Interest
Focal Interest

Science education

  1. GPT-4o shows partial understanding of science and GenAI-influenced science
    Study examines GPT-4o's capacity to represent the nature of science and epistemic dimensions of AI-influenced scientific knowledge production for educational contexts.
  2. Contextual STEM instruction showed preliminary gains in problem solving and motivation
    Quasi-experimental study examining culturally contextualized STEM instruction via the Engineering Design Process on problem-solving and motivation in Omani fourth-graders.
  3. GenAI shows mixed effects in computer science learning
    Systematic review of 64 studies examining generative AI's impact on computer science learning outcomes, hallucinations, and problem-solving skills across programming education contexts.
  4. GenAI matched human scoring of students’ scientific inquiry understanding
    Study evaluates generative AI's capacity to assess and provide feedback on elementary students' epistemic understanding in science inquiry, demonstrating high reliability with human raters.
  5. Misconceptions were linked to lower organic chemistry self-efficacy
    Study reveals that foundational chemistry misconceptions significantly predict lower self-efficacy in college organic chemistry, with implications for early intervention strategies.
  6. Interest-targeted course introductions improved data science scores
    Case study of data science education examining effects of interest-targeted instructional videos and self-directed data analysis exercises on student learning outcomes across 8,509 participants.
  7. Primary school robotics project showed varied student engagement
    Study examining situational engagement of fourth-grade students in robotics and coding project-based learning, identifying four distinct engagement profiles and optimal learning moments.
  8. Formative assessment data predicted standardized assessment performance
    Study examines predictive validity of computer-based formative assessment for standardised test outcomes, finding models explain 30-48% of variance with domain-specific patterns.