AI evidence extraction

Artificial intelligence in EFL higher education: effects on academic performance and social competence

Baderaddin Yassin, Nusaiba Ali Almousa, Asma Ali Almousa · Frontiers in Education · 2026

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

5/10

Relevance

2/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.3389/feduc.2026.1736757

Methodology & findings

Study design

Quasi-experimental pre-test/post-test control group design with 60 undergraduate EFL students assigned to either an AI-enhanced learning group (n=30) or conventional instruction group (n=30) over a 12-week intervention period.

Sample

N = 60, 4 groups

Primary method

Paired-sample t-tests were used to compare pre-test and post-test means within each group. Analysis of Covariance (ANCOVA) was used to compare post-test performance between groups while controlling for pre-test performance. SPSS or similar statistical software is implied but not explicitly stated.

Main result

The AI-enhanced group demonstrated statistically significant improvements in academic performance, with "a 15.4% increase compared to 6.9% in the control group." However, "differences in social competence gains between groups were not statistically significant." The study found that "AI-assisted instruction demonstrated significant advances in the academic achievement of EFL students" but that "AI tools will not always be concerned with the requirements of interpersonal skills, such as emotional expressivity, sensitivity and social control."

Reports effect sizes and confidence intervals.

Research paradigm

Positivist/Quantitative empirical research

Author conclusions

"This study examined how the AI-enhanced instruction influences academic achievement and social competence among undergraduate EFL students. The findings reveal that AI-assisted instruction demonstrated significant advances in the academic achievement of EFL students, which is in consistency with the prior research that supports the potential of AI in supporting the higher-order cognitive skills, academic literacy, and critical thinking." The authors conclude that "AI integration should be implemented within structured, blended learning environments aligned with course objectives. AI tools function most effectively as complements to instructor guidance rather than replacements. To foster social competence, pedagogical designs should intentionally incorporate collaborative and interaction-focused activities."

Risk of bias

Selection bias: quasi-experimental design without randomization; Hawthorne effect: novelty-related motivation in AI group; Construct-measurement misalignment for social competence; Confounding from multiple AI tools making isolation difficult; Instructional interaction differences between groups despite same instructor; Potential attrition not explicitly addressed; Hawthorne effect/novelty-related motivation potentially inflating AI group engagement; Inability to isolate individual AI tool contributions due to bundled intervention; Potential instructor bias in differential feedback despite same instructor teaching both groups; Measurement misalignment between SSI (personality-adjacent constructs) and language learning outcomes; Selection bias not addressed in quasi-experimental design (no randomization mentioned); No mention of blinding of assessors or participants; Selection bias: Quasi-experimental design without random assignment; Hawthorne effect: Novelty-related motivation in AI-enhanced group; Measurement bias: SSI may not adequately capture language-specific social competence; Confounding variables: Instructor effects, differences in instructional interaction and feedback associated with AI integration; Attrition risk: No mention of dropout rates or participant attrition; Construct-measurement misalignment: SSI measures personality-adjacent traits not directly related to language learning outcomes

Limitations

"Interpretation is limited in five ways
First, quick changes in models imply that any observed impact is in part moving target dependent." The authors acknowledge that "the intervention combined multiple AI tools targeting different skills, making it difficult to isolate the contribution of individual components." Additionally, "novelty-related motivation (i.e., a Hawthorne effect) may have temporarily increased engagement." The authors note that "the adapted Social Skills Inventory (SSI) demonstrated acceptable reliability, it primarily measures broad socio-emotional traits (e.g., expressivity, sensitivity, control) that are relatively stable and only indirectly related to language learning contexts." The study was limited to a single institution and 12-week duration.

Open questions raised

Limited population-specific studies on undergraduate EFL learners
Relative absence of longitudinal studies to establish long-term effects of AI usage
Lack of measurement instruments that could analyze social competence in advanced academic settings
Need for long-term studies examining whether AI contributes to scaffolding internalization or metacognitive offloading
Cross-cultural applicability of AI integration in EFL contexts
Limited studies specifically targeting undergraduate EFL students (most literature focuses on general language proficiency or graduate students)

Extracted from: pdfAgreement 55%

Explore related topics

Productivity Effects Research Productivity Experimental