aceModerate evidence
More
Review of 16 studies finds AI helps university learning, but evidence varies widely
Study title: University study programs in the AI era examined through a PRISMA guided meta analysis of global trends and regional disparities in the Western Balkans
- Evidence type
- Systematic review or meta-analysis
- Sample
- n = 16
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-2807cite or check it
Study title: University study programs in the AI era examined through a PRISMA guided meta analysis of global trends and regional disparities in the Western Balkans
A PRISMA-guided meta-analysis looked at 20 studies on AI in higher education, of which 16 met the eligibility requirements. Using random-effects meta-regression, it found a moderate, statistically significant positive effect, with a pooled Hedges' g of 0.685 (95% CI 0.389 to 0.982).
Read the study abstract
Artificial intelligence (AI) in higher education is becoming very important research field because of its extensive use, the absence of a thorough framework governing its use, ethical issues, and effects on student learning outcomes. In light of the increasing global commitment to incorporating AI into higher educational institutions, this study synthesizes current empirical findings using the PRISMA 2020 meta-analysis framework and estimate further the impact of AI in higher education using Meta Regression Anal…
The gain
A meta-analysis of 16 studies found that integrating AI in higher education had a moderate positive effect on learning outcomes (Hedges' g = 0.685, 95% CI 0.389 to 0.982).
In brief
The studies differed greatly: Cochran's Q was 337.02 (p < 0.001), I² was 95.5% and τ² was 0.3028. The authors attribute this to variation in research designs, target populations and ways AI was implemented in teaching, so the average effect hides large differences between settings.
The research base is geographically uneven, which limits how far the results can be generalised. The authors call for regionally diverse research, especially in Europe and the Western Balkans, and say national research is needed to map how AI is used across fields of study and what it does to learning outcomes.
Main points
- The meta-analysis followed the PRISMA 2020 framework and started from 20 studies; 16 met the eligibility requirements.
- The pooled effect of AI integration on higher-education outcomes was Hedges' g = 0.685, with a 95% confidence interval of 0.389 to 0.982.
- Heterogeneity was very high: Cochran's Q = 337.02 (p < 0.001), I² = 95.5%, τ² = 0.3028.
- The authors used meta-regression with restricted maximum likelihood (REML) and a random-effects model.
- The published evidence is geographically imbalanced, which restricts generalisability.
- The authors say little research exists on AI's use and effects in academic faculties, especially in Europe and the Western Balkans.
What the review found when it combined the studies
These are the review’s own pooled results, copied from its abstract. Truvace did not recalculate them.
Student learning outcomes with AI integration in higher education
Hedges' g 0.685 (95% CI 0.389 to 0.982)
16 studies
The review’s own words
The findings indicated a moderate and statistically significant positive impact of integrating AI in higher education, with a pooled effect size of Hedges’ g = 0.685 and a 95% confidence interval of [0.389, 0.982].
Code checks every number above against the review’s text before it is shown. The short labels are an AI model’s reading of that text, so the review’s own sentence sits under each one. A pooled figure is only as strong as the studies behind it. How Truvace uses pooled results
The full story
The study was designed as a systematic synthesis of empirical findings on AI in higher education, guided by PRISMA 2020. Of 20 studies identified, 16 satisfied all eligibility requirements and were pooled with meta-regression using restricted maximum likelihood.
The pooled result, Hedges' g of 0.685, indicates a moderate positive effect on learning outcomes. The wide confidence interval (0.389 to 0.982) shows the size of the benefit is uncertain, and an I² of 95.5% means nearly all variation reflects real differences between studies rather than chance.
The authors say this variation reflects differences in research designs, target populations and AI implementation strategies. They argue that the geographic imbalance of the literature restricts generalisability and call for strategic national research in the Western Balkans to map AI use across fields of study and assess its educational consequences.
Sources
1 on file- Discover Educationdoi:10.1007/s44217-026-01399-9
Talk it over
Who gets this benefit, and who was left out of the study or story?
Review of 16 studies finds AI helps university learning, but evidence varies widelyhttps://truvace.com/item/university-study-programs-in-the-ai-era-examined-through-a-prisma-guided-meta-an
Who gets this benefit, and who was left out of the study or story?
The gain A meta-analysis of 16 studies found that integrating AI in higher education had a moderate positive effect on learning outcomes (Hedges' g = 0.685, 95% CI 0.389 to 0.982).
Answer in the threadHow do we know? What can this kind of evidence show, and what can’t it?
The evidence This rests on a systematic review or meta-analysis with a sample of 16, from Discover Education.
Answer in the threadWhat does it leave open? What else would you want to know before acting on it?
The record’s own caveat Heterogeneity across the pooled studies was very high (I² = 95.5%), and the geographic imbalance of the literature restricts how far the findings generalise, particularly to Europe and the Western Balkans.
Answer in the threadHave you seen AI used this way where you study, work or live? Did it go the same way?
Answer in the thread
The questions are the same on every record. The details under them are taken from this record as written.
What readers say