The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2766
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T17:20:47.769299Z
status: published
lens: g_space
sector: education
headline: Student–AI interaction: unveiling the impact of AI agents on vocabulary learning – A meta-analysis
dek: Abstract The rapid proliferation of artificial intelligence (AI)-driven educational interventions has generated substantial interest in their effectiveness for vocabulary acquisition; however, the existing literature presents inconsistent findings. This meta-analysis, drawing on 28 studies and integrating 55 effect sizes via a three-level random effects model, demonstrates a significant overall positive effect of AI agents on vocabulary learning (effect size = 0.904, 95 % CI: 0.724, 1.084, p < 0.0001). Furthermo…
gain_title: A meta-analysis of 28 studies and 55 effect sizes found AI agents had a significant positive effect on vocabulary learning (effect size 0.904, 95% CI 0.724 to 1.084).
problem_title: (none)
trace_subject: (none)
gain_reading: A meta-analysis of 28 studies and 55 effect sizes found AI agents had a significant positive effect on vocabulary learning (effect size 0.904, 95% CI 0.724 to 1.084).
gain_evidence: demonstrates a significant overall positive effect of AI agents on vocabulary learning (effect size = 0.904, 95 % CI: 0.724, 1.084, p < 0.0001)
problem_reading: (none)
problem_evidence: (none)
quick_read: A meta-analysis pooling 28 studies and 55 effect sizes found that AI agents had a significant positive effect on vocabulary learning, with an overall effect size of 0.904 (95% CI 0.724 to 1.084).

The authors used a three-level random effects model, which accounts for several effect sizes coming from the same study. They set out to resolve inconsistent findings in earlier research on AI-driven educational interventions for vocabulary acquisition.

Moderator analysis found that both the degree of AI embodiment and the duration of the intervention significantly shaped the size of the effect, so results varied with how the AI agent was designed and how long students used it.
limitation: 
tag: Evidence-backed gain
key_points: The meta-analysis covers 28 studies and 55 effect sizes on AI agents and vocabulary learning. | The pooled effect size was 0.904 (95% CI 0.724 to 1.084, p < 0.0001). | A three-level random effects model was used to combine the effect sizes. | AI embodiment was a significant moderator (F(1, 53) = 7.446, p < 0.01). | Intervention duration was also a significant moderator (F(20, 24) = 5.563, p < 0.001). | The analysis draws on a student-AI interaction framework to interpret instructional designs.
rundown: The review was motivated by the rapid spread of AI-driven educational interventions and by inconsistent findings in the existing literature on their effect on vocabulary acquisition.

Researchers combined 55 effect sizes from 28 studies using a three-level random effects model and reported a significant overall positive effect of 0.904, with a confidence interval of 0.724 to 1.084.

Moderator tests showed that AI embodiment and intervention duration both significantly affected outcomes. The authors frame the findings within the student-AI interaction framework as a theory-informed synthesis of how AI-supported instructional designs relate to vocabulary learning.
sources:
- peer_reviewed | International Review of Applied Linguistics in Language Teaching | https://doi.org/10.1515/iral-2025-0382 | 2026-08-12
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

5b127d6193beb0bd5de2ddc3d1f1310e577d7b2ff5ba5b76700cd589c0ba86a7

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000