Testing Theories of Transfer Using Error Rate Learning Curves.

Area of Science:

• Cognitive Science
• Educational Technology
• Learning Sciences

Background:

• The scope of learning transfer, or applying knowledge to new tasks, is a fundamental question in education.
• Existing theories propose either broad transfer (general applicability) or component transfer (specific skill application).

Purpose of the Study:

• To statistically model and compare broad transfer and component transfer theories using real-world student data.
• To evaluate which theoretical model better explains task difficulty and learning transfer in educational technology use.

Main Methods:

• Developed statistical models with latent variables representing cognitive functions impacting error rates.
• Contrasted strong models (common explanation for difficulty and transfer) with weak models (decoupled explanations).
• Evaluated models on prediction accuracy and explanatory power across eight student datasets.

Main Results:

• Component transfer models demonstrated superior prediction accuracy and explanatory power compared to broad transfer models.
• Weak model variations improved student generalization but reduced item generalization and explanatory power.
• The study identified specific, malleable cognitive components, such as spatial reasoning and executive functions.

Conclusions:

• Component theories provide a more accurate framework for understanding learning transfer than broad theories.
• Statistical modeling of educational technology use can reveal underlying cognitive mechanisms.
• This approach can guide the identification of specific cognitive skills for targeted educational interventions.