Impact of Chatbots on Chinese Speaking Skills Among International Students in China and Its Mechanism

This research conducted an 8-week randomized controlled trial to investigate the impact of the AI chatbot "Doubao" on Chinese oral proficiency among international students in China and explore its underlying mechanisms. The experimental group (30 participants) engaged in high-frequency interactive dialogue training using "Doubao," while the control group (30 participants) performed traditional oral exercises. Results showed that the experimental group achieved significant improvements in vocabulary (mean improvement of 3.3 points, p &lt; 0.001) and meaningful expression (2.8 points, p &lt; 0.001), with a total score increase of 12.2 points—substantially exceeding the control group's 3.4-point gain (p &lt; 0.001). Large effect sizes were observed for vocabulary (Cohen’s d = 1.1) and meaningful expression (d = 1.22). ANCOVA analysis confirmed the experimental group's superiority in fluency and syntax (e.g., F = 36.42, partial η² = 0.39), though pronunciation improvement remained limited (η² = 0.06), likely due to speech recognition technology constraints. Textual analysis revealed three core mechanisms: contextualized input (e.g., simulated shopping/medical scenarios enhancing vocabulary application), immediate feedback internalization (29.3% increase in grammatical error correction rates), and emotional incentives (68.4% learner-initiated topic rate). A case study of Student P23 demonstrated substantial growth in linguistic complexity (50.8% increase in complex sentence usage) and error correction rates (from 63.5% to 82.1%). The findings validate that AI chatbots address critical gaps in traditional pedagogy—such as authentic language contexts and personalized guidance—through high-frequency interactions and precision feedback, offering empirical support for digital transformation in international Chinese education. The study also highlights the need to optimize pronunciation training modules for comprehensive oral proficiency enhancement.本研究开展了一项为期8周的随机对照试验，旨在探究AI聊天机器人"豆包(Doubao)"对在华国际学生中文口语能力的影响，并揭示其背后的作用机制。实验组共30名参与者，使用豆包开展高频互动对话训练；对照组同样包含30名参与者，仅进行传统口语练习。结果显示，实验组在词汇量（平均提升3.3分，p<0.001）与有效表达（平均提升2.8分，p<0.001）维度均取得显著进步，总分提升达12.2分，远高于对照组的3.4分提升幅度（p<0.001）。词汇量（科恩d值(Cohen’s d)=1.1）与有效表达（d=1.22）的效应量均较大。协方差分析(ANCOVA)结果证实，实验组在流利度与句法维度表现更优（如F=36.42，偏η²=0.39），但发音提升效果有限（η²=0.06），这大概率源于语音识别技术的局限性。 文本分析揭示了三大核心作用机制：其一为情境化输入，例如模拟购物、就医场景以强化词汇应用能力；其二为即时反馈内化，该模式下语法纠错率提升29.3%；其三为情感激励，学习者主动发起话题的比例达68.4%。对学生P23的个案研究显示，其语言复杂度（复杂句使用量提升50.8%）与语法纠错率（从63.5%升至82.1%）均实现大幅增长。 本研究结果证实，AI聊天机器人可通过高频互动与精准反馈，弥补传统教学在真实语言情境与个性化指导方面的关键短板，为国际中文教育的数字化转型提供了实证支撑。同时，本研究也指出需进一步优化发音训练模块，以实现口语能力的全面提升。

搜集汇总

背景与挑战

背景概述

该数据集基于一项随机对照试验，研究AI聊天机器人'Doubao'对中国国际学生汉语口语技能的影响及其机制。研究发现，实验组通过高频互动和精准反馈，在词汇和表达方面取得显著提升，但发音改善有限，揭示了AI在语言教育中的优势和局限性。数据集包含实验组和对照组的得分数据，支持语言教育领域的数字化转型研究。