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-2026-2724
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T17:14:59.543001Z
status: published
lens: g_space
sector: education
headline: The effects of AI-guided individualized language learning: A meta-analysis
dek: Artificial intelligence (AI) has considerably advanced the methods for individualizing language learning opportunities, such as assessing learning progress and recommending effective individual instruction. In the present study, we conducted a meta-analysis to synthesize recent empirical findings pertaining to the utilization of AI-guided language learning and collected 61 samples (N = 8,282) from 17 research projects (e.g., Assessment to Instruction [A2i], Duolingo, and Project LISTEN). The results of our meta-…
gain_title: A meta-analysis of 61 samples (N = 8,282) from 17 research projects found AI-guided individualized language learning beat business-as-usual instruction by d = 0.39 across 35 between-group samples (N = 6,020).
problem_title: (none)
trace_subject: (none)
gain_reading: A meta-analysis of 61 samples (N = 8,282) from 17 research projects found AI-guided individualized language learning beat business-as-usual instruction by d = 0.39 across 35 between-group samples (N = 6,020).
gain_evidence: had an overall positive treatment effect compared to business-as-usual conditions (d = 0.39, based on 35 between-group samples, N = 6,020) | AI-guided individualized language learning was effective for learners’ language development (d = 1.18, based on 26 within-group samples, N = 2,262)
problem_reading: (none)
problem_evidence: (none)
quick_read: A meta-analysis pooled 61 samples covering 8,282 learners from 17 research projects, including Assessment to Instruction (A2i), Duolingo and Project LISTEN, to test AI-guided individualized language learning.

Across 26 within-group samples (N = 2,262), learners improved substantially over time (d = 1.18). Across 35 between-group samples (N = 6,020), AI-guided learning outperformed business-as-usual conditions by a smaller margin (d = 0.39).

Moderator analyses found that systems built with machine learning and hybrid approaches had larger effects than rule-based systems. The authors note that rule-based systems may be easier to interpret pedagogically, since it is clearer how their predictions are made.
limitation: 
tag: Evidence-backed gain
key_points: The analysis collected 61 samples with 8,282 learners from 17 research projects. | Projects included Assessment to Instruction (A2i), Duolingo and Project LISTEN. | Within-group samples (26, N = 2,262) showed a large gain in language development, d = 1.18. | Between-group samples (35, N = 6,020) showed a positive effect over business-as-usual teaching, d = 0.39. | Machine learning and hybrid systems had larger effects than rule-based systems. | Rule-based systems may make it easier to understand how predictions are made from a teaching perspective.
rundown: The meta-analysis synthesized recent empirical work on AI that assesses learning progress and recommends individual instruction for language learners, drawing on 17 research projects.

Two kinds of comparison were pooled. Within-group designs, which track learners' progress over time, gave an effect of d = 1.18 across 26 samples. Between-group designs, which compare AI-guided learners with business-as-usual conditions, gave a smaller effect of d = 0.39 across 35 samples.

Moderator analyses showed machine learning and hybrid systems were more impactful than rule-based ones, though the authors say rule-based systems may be more useful for understanding how predictions are made pedagogically.
sources:
- peer_reviewed | Language Learning & Technology | https://doi.org/10.64152/10125/73575 | 2024-06-01
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

4e617faa9aad8d4a5def4f961752232b86dd0d4e4db17047fa38b8d2eb70b0ea

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000