Truace
Subscribe
70/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Nine studies of AI-supported intercultural lessons show large learning gains

Study title: Emerging Technology-Supported Intercultural Learning for Cognitive, Attitudinal, and Behavioral Outcomes: A Systematic Review and Three-Level Meta-Analysis

Evidence type
Systematic review or meta-analysis
Sources
1 sourcePeer-reviewed
Published
3 min read
Truvace record
TRV-2026-2708cite or check it

Study title: Emerging Technology-Supported Intercultural Learning for Cognitive, Attitudinal, and Behavioral Outcomes: A Systematic Review and Three-Level Meta-Analysis

A systematic review and three-level meta-analysis pooled 17 studies published from January 2020 to July 2026 (77 effect sizes) on technology-supported intercultural learning. Across AI, augmented reality and virtual reality, learners beat comparison groups with a pooled Hedges' g of 0.965. The nine AI-supported studies, which used chatbots, generative AI, intelligent agents and adaptive systems, gave a pooled g of 1.150 across 43 effect sizes.

Read the study abstract

Emerging technologies are increasingly used to support intercultural learning, yet evidence on the distinct effects of artificial intelligence (AI), augmented reality (AR), and virtual reality (VR) remains limited. This systematic review and three-level meta-analysis synthesized 17 studies published from 2020 to July 2026 (77 effect sizes) to examine cognitive, attitudinal, and behavioral outcomes. The overall effect was large and significant (Hedges’ g = 0.965, 95% CI [0.580, 1.350], p < .001), although heterog…

The gain

In a meta-analysis of nine controlled studies (43 effect sizes), learners taught intercultural skills with AI tools such as chatbots and adaptive systems scored higher than comparison groups, with a pooled effect of g = 1.150 (95% CI 0.744 to 1.555).

In brief

The AI studies covered different tools and tasks, including translation and feedback in a 12-week online exchange between Chinese and Malaysian students, an adaptive system that flagged culturally inappropriate expressions, and AI responses with cultural bias that learners were asked to critique. The authors say gains often relied on teacher-designed tasks, structured reflection and human partners, so the effect cannot be credited to the AI alone.

Results varied widely between studies. The prediction interval for the AI studies ran from about -0.29 to 2.59, meaning a new comparable study could find little or no benefit. All randomized studies in the review had some risk-of-bias concerns, and the evidence base is only 17 studies.

Main points

  1. The review pooled 17 studies (77 effect sizes) published between January 2020 and July 2026; eight were published in 2026.
  2. The overall pooled effect of AI, AR and VR on intercultural learning was g = 0.965 (95% CI 0.580 to 1.350), with a prediction interval of about -0.781 to 2.711.
  3. Nine AI-supported studies contributed 43 effect sizes and a pooled g of 1.150; dropping the one preprint raised it slightly to g = 1.217 across eight studies.
  4. Heterogeneity was very high (total I-squared 93.88%), and most of it came from differences between studies rather than within them.
  5. Outcome type (cognitive, attitudinal or behavioral) did not significantly change the effect; cognitive outcomes had g = 0.892.
  6. AR looked stronger than VR (two AR studies versus six VR studies), but the gap was no longer significant after excluding one AR study.
  7. K-12 and larger-sample studies showed bigger effects, but only two studies involved K-12 learners and the authors call these comparisons exploratory.
  8. Three of the 17 studies compared against technology-based alternatives; the other 14 used conventional or non-technology instruction.

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.

  • Intercultural learning outcomes among adult learners with emerging technology compared with control conditions

    Hedges' g 0.758 (95% CI 0.469 to 1.047)

    15 studies · search to July 2026

    The review’s own words
    Adult learners were used as the reference group and showed a positive estimated effect, g = 0.758, 95% CI [0.469, 1.047].

  • Cognitive intercultural learning outcomes with emerging technology compared with control conditions

    Hedges' g 0.892 (95% CI 0.469 to 1.315)

    14 studies · search to July 2026

    The review’s own words
    The cognitive reference group showed a significant positive effect, g = 0.892, 95% CI [0.469, 1.315], p < .001.

  • Intercultural learning outcomes with AI-, AR- or VR-supported interventions compared with control conditions

    Hedges' g 0.965 (95% CI 0.580 to 1.350)

    17 studies · search to July 2026

    The review’s own words
    The pooled effect was statistically significant, g = 0.965, 95% CI [0.580, 1.350], t (76) = 4.997, p < .001.

  • Difference in intercultural learning effect between virtual reality and augmented reality interventions

    Beta (difference in Hedges' g) −1.481 (95% CI −2.681 to −0.280)

    search to July 2026

    The review’s own words
    Compared with AR, VR-based interventions showed a significantly smaller estimated effect, β = −1.481, 95% CI [−2.681, −0.280], p = .017.

  • Intercultural learning outcomes with AI-supported interventions compared with control conditions

    Hedges' g 1.150 (95% CI 0.744 to 1.555)

    search to July 2026

    The review’s own words
    AI-supported interventions showed a significant positive effect, g = 1.150, 95% CI [0.744, 1.555], p < .001.

  • Intercultural learning outcomes with augmented reality interventions (reference group in AR vs VR comparison)

    Hedges' g 1.856 (95% CI 0.820 to 2.893)

    search to July 2026

    The review’s own words
    AR was used as the reference group and showed a positive estimated effect, g = 1.856, 95% CI [0.820, 2.893], p < .001.

  • Intercultural learning outcomes in large-sample studies with emerging technology compared with control conditions

    Hedges' g 1.575 (95% CI 1.029 to 2.121)

    search to July 2026

    The review’s own words
    This group showed a significant positive effect, g = 1.575, 95% CI [1.029, 2.121], p < .001.

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 authors searched Web of Science, Scopus and ProQuest through 31 July 2026 and included only randomized or quasi-experimental studies with a comparison group and enough data to calculate effect sizes. Two reviewers screened, coded and assessed risk of bias independently, with high agreement on coding (Cohen's kappa 0.96). Outcomes were sorted into cognitive, attitudinal and behavioral domains.

For AI, the studies used tools for different jobs: translation and feedback in a 12-week Chinese-Malaysian exchange, an adaptive system that detected and explained culturally inappropriate wording, an AI agent that responded during pragmatic conflicts, and deliberately biased AI answers used for critique. The largest effects in two studies appeared in critical cultural awareness and metacognitive cultural intelligence rather than factual knowledge.

Quality checks gave a mixed picture. All randomized studies were rated as having some concerns, five of seven non-randomized studies had moderate risk, and two were rated serious. Excluding those two serious-risk studies raised the overall effect to g = 1.110. Funnel-plot, Egger's and Begg's tests showed no strong sign of publication bias, though 17 studies may be too few to detect it.

Moderator findings were fragile. Academic level and sample size were significant, but K-12 was represented by two studies, sample size was unevenly spread across technology types, and the covariance-based sensitivity checks covered only the overall effect, not the moderators.

Sources

1 on file
  1. Journal of Intelligencedoi:10.3390/jintelligence14100265
    Peer-reviewed2026-10-16

Talk it over

Four questions to start onWho gets this benefit, and who was left out of the study or story?

Start with a question

Nine studies of AI-supported intercultural lessons show large learning gainshttps://truvace.com/item/emerging-technology-supported-intercultural-learning-for-cognitive-attitudinal-a

  1. Who gets this benefit, and who was left out of the study or story?

    The gain In a meta-analysis of nine controlled studies (43 effect sizes), learners taught intercultural skills with AI tools such as chatbots and adaptive systems scored higher than comparison groups, with a pooled effect of g = 1.150 (95% CI 0.744 to 1.555).

    Answer in the thread

  2. How 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, from Journal of Intelligence.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat Effects varied widely between studies (the AI prediction interval was about -0.29 to 2.59), the review covers only 17 studies, all randomized studies had some risk-of-bias concerns, and the moderator comparisons were exploratory with uneven groups.

    Answer in the thread

  4. Have 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

More from Education