Four-category operationalization of Winne and Hadwin's four-stage SRL model for tutored problem-solving verbalizations¶

Theory · All theories
Evidence · no claims cited

Description¶

The study operationalizes SRL in think-aloud transcripts as four categories grounded in Winne and Hadwin's four-stage model: "processing information, planning, enacting, and realizing errors". Concatenated utterances between consecutive tutor transactions were annotated with a coding scheme (Table 1) specifying behaviors per category, such as assembling information, forming plans, verbalizing actions, and realizing mistakes. This subset of stage-relevant behaviors enables relating verbalized SRL to tutor action correctness.

Design Implications¶

Context¶

Requirements¶

• Think-aloud utterances segmented and synchronized with timestamped tutor transactions (no more than a 1-second error margin)
• Trained human coders establishing inter-rater reliability before coding the full sample

Constraints¶

• Categories represent a subset of SRL processes within each model stage, focusing on problem-solving learning environments

Target Learners¶

• university students in chemistry and formal logic courses

Target Learning Objectives¶

• self-regulated learning during tutored problem-solving

Claims¶

Related Theories¶

• Self-Regulated Learning
• Three-phase SRL scaffolding framework (planning, task execution, post-task reflection) operationalized across iterative design cycles

Examples¶

• Multilingual SRL detection pipeline using OpenAI text-embedding-3-small embeddings with a single-hidden-layer neural network classifier

Key Sources¶

• Conrad Borchers, Jiayi Zhang, Hendrik Fleischer, Sascha Schanze, Vincent Aleven, Ryan S. Baker. (2025). Large Language Models Generalize SRL Prediction to New Languages Within But Not Between Domains. Journal of Educational Data Mining, Volume 17, No 2. https://github.com/pcla-code/EDM24_SRL-detectors-for-think-aloud