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-2026-2767
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T17:21:01.815462Z
status: published
lens: g_space
sector: education
headline: Multimodality and social interactions in AI-Enhanced embodied robot-assisted language learning: A meta-analysis
dek: Embodied robots are increasingly used to support second language (L2) learning, yet evidence on when and why they help remains mixed. This meta-analysis examines how robot design features and instructional conditions relate to L2 learning outcomes, drawing on multimodal learning theory and sociocultural theory. Following PRISMA 2020 guidelines, Web of Science and Scopus were searched for empirical studies that used physically embodied, AI-enhanced robots in L2 settings and reported sufficient statistics to compu…
gain_title: A meta-analysis of 11 studies (595 learners) found AI-enhanced embodied robots were linked to better second language outcomes, with a pooled effect of Hedges' g = 0.83 (95% CI 0.46 to 1.21).
problem_title: (none)
trace_subject: (none)
gain_reading: A meta-analysis of 11 studies (595 learners) found AI-enhanced embodied robots were linked to better second language outcomes, with a pooled effect of Hedges' g = 0.83 (95% CI 0.46 to 1.21).
gain_evidence: yielded a positive overall effect (Hedges’ g = 0.83, 95% CI [0.46, 1.21], p < .001) | Eleven studies (17 effect sizes, N = 595) met the inclusion criteria
problem_reading: (none)
problem_evidence: (none)
quick_read: A meta-analysis of 11 empirical studies covering 595 learners examined physically embodied, AI-enhanced robots in second language settings. It found a positive overall effect on learning outcomes, with Hedges' g of 0.83 (95% CI 0.46 to 1.21).

Of six design and instruction factors tested, only the robot-learner interaction format reached significance (p = .049): group-based formats showed larger effects than one-on-one interaction. Robots used as teaching assistants had the largest subgroup effect, though social role did not significantly moderate outcomes.

Richer multimodal and more autonomous designs were not consistently linked to stronger learning. The authors suggest outcomes related more to how robots are positioned within instruction than to technical sophistication, though heterogeneity was high and the evidence base is small.
limitation: Heterogeneity was high (I2 = 84.4%) and the evidence base was small, so the authors say the pooled estimate should be read with caution.
tag: Evidence-backed gain
key_points: The review followed PRISMA 2020 guidelines and searched Web of Science and Scopus for studies using physically embodied, AI-enhanced robots in second language settings. | Eleven studies with 17 effect sizes and 595 participants met the inclusion criteria. | A three-level random-effects model with cluster-robust variance estimation gave a pooled Hedges' g of 0.83 (95% CI 0.46 to 1.21, p < .001). | Heterogeneity was high (I2 = 84.4%), and the authors say the estimate should be read with caution. | Group-based interaction showed larger effects than one-on-one interaction (p = .049), the only significant moderator of six. | Robot morphology, communicative modality, autonomy, social role and intervention duration did not significantly moderate outcomes.
rundown: The analysis was framed by multimodal learning theory and sociocultural theory, and aimed to clarify when and why embodied robots help second language learners, given mixed earlier evidence.

Only studies that used physically embodied, AI-enhanced robots and reported enough statistics to compute effect sizes were included. The final set was 11 studies yielding 17 effect sizes among 595 learners.

The pooled effect was positive and statistically significant, but heterogeneity was high at 84.4%. Among six moderators, only interaction format reached significance, with group formats outperforming one-on-one use.

Robots in teaching-assistant roles showed the largest subgroup effect, but social role was not a significant moderator. More multimodal and more autonomous designs were not consistently tied to stronger learning, which the authors read as pointing to instructional positioning over technical sophistication.
sources:
- peer_reviewed | Educational Research Review | https://doi.org/10.1016/j.edurev.2026.100833 | 2026-08-01
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

a94023d2648b8fddbd3af4e7686396dee481dff3e114471514ae29dc55fe6e40

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000