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-2026-2747
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-12T17:18:20.964393Z
status: published
lens: g_space
sector: education
headline: THE EFFECTS OF CHATBOTS-BASED INTERVENTIONS ON FOREIGN LANGUAGE SPEAKING PERFORMANCE: A META-ANALYSIS STUDY
dek: The aim of this study is to examine the effect of chatbot-based interventions on foreign language speaking performance through a meta-analysis. Studies investigating the effects of chatbots on speaking performance were retrieved from various databases (e.g., ERIC, ScienceDirect, Taylor & Francis) using a range of keyword phrases. Following the inclusion and exclusion criteria, this meta-analysis included 16 experimental studies (k = 17) that implemented chatbot-based interventions to enhance learners’ foreign la…
gain_title: A meta-analysis of 16 experimental studies (17 effect sizes) found chatbot-based practice raised foreign-language speaking performance over control groups, with a pooled Hedges' g of 1.035 (95% CI 0.632 to 1.439), a large effect.
problem_title: (none)
trace_subject: (none)
gain_reading: A meta-analysis of 16 experimental studies (17 effect sizes) found chatbot-based practice raised foreign-language speaking performance over control groups, with a pooled Hedges' g of 1.035 (95% CI 0.632 to 1.439), a large effect.
gain_evidence: The overall effect size of the intervention studies on foreign language speaking performance was 1.035, indicating a large effect. | g = 1.035 (%95 CI [0.632, 1.439], p < .05)
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers at Trabzon University pooled 16 experimental studies (17 effect sizes) that tested chatbot-based interventions for foreign-language speaking. The pooled effect was a Hedges' g of 1.035, which the authors class as large, meaning chatbot groups outperformed control groups.

The studies came from countries including Taiwan, China, Jordan, Egypt, Iran, Korea and Turkiye, and covered primary pupils, high schoolers and university students. Sample sizes ranged from 42 to 93 per study. Results varied widely: 14 of 17 effect sizes were positive and 3 were negative, and heterogeneity was high (I² = 91.89).

One study, El-Said (2023), reported an outlying effect of 5.825. Without it, the pooled estimate fell to 0.793. Neither education level nor intervention duration differed significantly, though duration came close (p = 0.053), with long-term programmes averaging 1.864.
limitation: Heterogeneity across the studies was very high (I² = 91.89), the pooled effect was driven partly by an outlier, and the funnel plot showed asymmetry suggesting possible publication bias; only three high school studies were included.
tag: Evidence-backed gain
key_points: The review searched 12 databases, found 101 unique reports and kept 16 quasi- or true-experimental studies after excluding 29 that did not address speaking performance, 6 without usable statistics and 48 non-quasi-experimental studies. | Pooled effect: Hedges' g = 1.035 under a random-effects model; 11 of 17 effect sizes were statistically significant and 6 were not. | Removing the outlier study (El-Said, 2023, g = 5.825) lowered the pooled effect from 1.035 to 0.793. | Three effect sizes were negative or near zero: Chien et al. 2022 and Deng 2024 (both -0.008) and Wang et al. 2024's Typebot group (-0.086). | By education level, average effects were 1.225 for university students, 0.747 for K-8 and 0.265 for high school; the differences were not significant (p = 0.183). | By duration, average effects were 1.864 for long-term (9+ weeks), 0.603 for mid-term and 0.358 for short-term interventions; the difference was not significant (p = 0.053). | Mean methodological quality of the included studies was 12.9 out of 18 on the MERSQI scale.
rundown: The authors followed a PRISMA process, with two researchers searching databases between 15 and 30 January 2025 and coding the studies independently, reaching 94% agreement. Effects were computed as Hedges' g in Comprehensive Meta-Analysis software using a random-effects model.

The included trials used tools such as Duolingo, Replika, LINE, Mondly and TPBOT, with control groups taught by other methods. Sample sizes ran from 42 to 93 learners. Six effect sizes exceeded 1, five fell between 0.5 and 1 and six were below 0.5.

Publication bias checks gave mixed signals: Egger's test found no significant bias (p = .225), but the funnel plot was asymmetric, and a trim-and-fill adjustment suggested four missing studies and a 26% difference between observed and corrected values. Fail-safe N values were 951 and 101.

The authors say the field is still early. Chatbot type could not be tested as a moderator because too few studies existed, and speaking sub-skills such as fluency, accuracy and pronunciation were combined into one score.
sources:
- peer_reviewed | Turkish Online Journal of Distance Education | https://doi.org/10.17718/tojde.1796214 | 2026-10-01
prev: 0000000000000000000000000000000000000000000000000000000000000000

SHA-256 hash

8b7db970879e79f62028135c8a232d7440c6cdede8cae195ace6e499c946a52e

Previous version’s hash

0000000000000000000000000000000000000000000000000000000000000000