AI evidence extraction

When Generative AI Meets Socratic Method: Investigating Programming Learning Dynamics Through Behaviours, Interaction Qualities and Perceptions

Dan Sun, Yi Zheng, Zhanshan Yang · Journal of Computer Assisted Learning · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

6/10

Relevance

1/4

Quality (LMQS)

Evidence

Citations

74.11

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1002/jcal.70210

Methodology & findings

Study design

Quasi-experimental study comparing two pedagogical approaches (GAI-Scaffolded Learning vs.

Sample

N = 80, 2 groups

Primary method

Click stream analysis, lag-sequential analysis, epistemic network analysis (ENA), and statistics; specific statistical tests, software, or significance levels are not detailed in the abstract.

Main result

The study found that "GSL engaged in cyclical, reflective practices (debugging, Socratic questioning, console use), while GDL prioritised rapid fixes via trial‐and‐error with GPT code, risking superficial mimicry and over‐reliance on external resources." Additionally, "ENA highlighted GSL's deeper engagement through interconnected feedback, emotional support and iterative inquiry, reducing frustration and sustaining persistence and GDL interactions focused on surface‐level queries, lacking scaffolding for emotional/heuristic integration."

Reports effect sizes.

Research paradigm

Pragmatist/Mixed-methods (quantitative and qualitative)

Author conclusions

The authors conclude that "Based on these findings, the study proposes pedagogical and developmental implications for future design and development of AI‐augmented curricula, providing actionable insights for educators seeking to harness GAI's potential while nurturing critical thinking in programming education."

Risk of bias

Quasi-experimental design (not randomized) may introduce selection bias; Small sample size (n=80) limits generalizability; Self-selection bias in treatment assignment not mentioned; Potential confounding variables from individual differences in learning styles; Selection bias: Quasi-experimental design without randomization may introduce systematic differences between groups; Hawthorne effect: Screen recording and observation may alter participant behavior; Potential confounding variables related to learner prior knowledge or aptitude not explicitly controlled; Single institution sample may limit generalizability; Potential selection bias in student assignment to conditions; No explicit mention of randomization in quasi-experimental design; No mention of blinding of instructors or participants; Potential attrition not addressed in abstract

Open questions raised

The study addresses the gap that "The integration of generative artificial intelligence (GAI) tools like GPT into programming education offers transformative potential through personalised guidance and instant feedback, yet risks fostering overreliance and superficial learning due to their tendency to deliver direct, context‐free answers."
The study identifies gaps in understanding how to optimize GAI-facilitated programming instruction to emphasize critical thinking over passive solution retrieval, and proposes future design and development of AI-augmented curricula.
The study addresses a gap in understanding how generative AI can be optimized for programming education through Socratic methods to avoid overreliance and superficial learning. Future research directions include design and development of AI-augmented curricula informed by these findings.

Data: not_statedCode: not_statedExtracted from: pdfAgreement 64%

Explore related topics

AI Literacy Training Doctoral Training