Hello everyone. This is Li, a graduate student. Here, I would like to introduce a paper I read at the recent English seminar.

Title: Unpacking AI-supported Chinese as a foreign language learning: How beginner-level learners’ cognitive and motivational factors predict speaking proficiency
• Authors: Jie Zhang, Xiaosheng Zhou, Ying Soon Goh
• Journal: Acta Psychologica (2025)
• DOI: https://doi.org/10.1016/j.actpsy.2025.105703

1. Research Background and Theoretical Framework
This paper investigates how the perceptions and AI usage habits of beginner-level learners relate to their speaking proficiency in AI-supported Chinese language learning.

In recent years, generative AI such as ChatGPT has been rapidly integrated into foreign language learning. Using AI enables various learning supports, including creating example sentences, translation, explaining vocabulary, pronunciation practice, script writing, and conversation practice. However, using AI does not automatically lead to higher learning outcomes. What is important is how learners perceive and utilize AI.

This study employs the Technology Acceptance Model (TAM). This is a model that explains how people evaluate and accept new technology when they use it. This study focuses on four specific factors: “perceived usefulness of AI,” “self-efficacy in using AI,” “trust in AI,” and “positive attitude toward using AI.”

Furthermore, the learning activities in this study focus on actual expression in Chinese rather than mere knowledge verification. Learners work in groups to create role-play scripts and present them in Chinese. Therefore, how AI relates not only to script writing but also to actual speaking proficiency is a critical issue.

2. Research Questions
This study sets out three main research questions.

First, to clarify how trust in AI, perceived usefulness, self-efficacy, and positive attitudes toward AI usage relate to Chinese speaking proficiency.
Second, to examine which of these factors has the strongest relationship with speaking proficiency.
Third, to investigate how AI usage relates to performance in script writing and oral presentation, respectively.

In other words, the focus of this study is not simply “whether AI was used.” Rather, it attempts to clarify how learners thought about and used AI, and how those usage patterns connect to their Chinese presentation abilities.

3. Research Methods
This study employs both quantitative analysis and qualitative analysis, such as interviews. First, it uses survey and grade data to examine the relationship between learners’ perceptions of AI and their Chinese language performance. Then, it uses interviews, scripts created by learners, and teacher observation records to examine the learning process in detail, which cannot be captured by numbers alone.

Participants were 120 beginner-level students learning Chinese at a university in Malaysia. All participants were non-native learners of Chinese and had experience using some form of AI language learning tool.

In class, AI-supported vocabulary practice, oral practice, and conversation activities were conducted from the first to the 12th week. From the 13th to the 14th week, students created role-play scripts in groups, and from the 15th to the 16th week, they presented in Chinese based on those scripts. Finally, a survey on AI usage experience was conducted.

Learners were not required to use specific AI tools. They used ChatGPT, translation apps, pronunciation feedback tools, and AI writing tools based on their own judgment.

Chinese oral proficiency was evaluated based on the presentations. Evaluation criteria included pronunciation, intonation, fluency, accuracy of vocabulary and grammar, expressiveness, and task achievement. Final grades were based on a 75% weight for oral presentation and 25% for script writing.

4. Main Results
Analysis results showed that all four factors related to AI were associated with Chinese speaking proficiency. Particularly important was the self-efficacy in using AI. Learners who could use AI effectively showed higher grades in oral presentations.

High-achieving learners did not use AI-generated answers as they were; instead, they checked the content themselves, revised it as necessary, and utilized it in a way that suited their own level. In other words, they used AI not as a “machine that provides answers” but as a “tool that supports learning.”

Also, learners who were positive about using AI and those who felt that AI was useful for Chinese learning tended to engage in tasks more actively. As a result, higher learning outcomes were observed.

On the other hand, regarding trust in AI, slightly unexpected results were shown. Learners who trusted AI too much did not necessarily show high outcomes. This is thought to be because assuming AI’s answers are correct reduces opportunities to think or verify information on their own.

Furthermore, AI usage was strongly related to script writing performance. AI was very helpful for script writing as it assisted with word choice, sentence structure, and content organization. However, the relationship between AI usage and actual oral presentation performance was not as strong. This means that the ability to write a good script and the ability to actually speak in Chinese are not necessarily the same.

5. Results of Qualitative Analysis
Interviews and teacher observations revealed several characteristics in learners’ AI usage.

First, high-achieving learners viewed AI answers critically. Instead of using AI-generated sentences as they were, they checked the meaning and revised them into expressions they could speak themselves. Such learners tried to use Chinese as their own language in the end, while utilizing AI skillfully.

On the other hand, some low-achieving learners simply copied sentences created by AI. In such cases, even if the script looked good at first glance, problems occurred where the individual did not fully understand the content or used words they could not pronounce.

Also, expressions created by AI were sometimes too difficult for beginner learners. Memorizing difficult sentences as they are leads to unnatural presentations or an inability to answer impromptu questions. From this perspective as well, it is necessary to adjust AI output to the learner’s level rather than accepting it as is.

6. Discussion
The following is the authors’ discussion. An important point learned from this study is that while AI is a convenient tool to support foreign language learning, its effectiveness varies greatly depending on how it is used. Learners who use AI effectively used it not as a “machine that provides the correct answer” but as a “tool to support learning.” It was observed that while they used AI to get ideas or check vocabulary and sentence patterns, they ultimately revised them into their own words.

On the other hand, relying too much on AI may prevent the development of independent thinking and expression skills. Especially in speaking learning, just writing a script is insufficient. The authors point out the necessity of actually speaking aloud, checking pronunciation, and practicing answering impromptu questions.

Therefore, while AI can assist in speaking learning, it is considered difficult to improve speaking skills with AI alone. They suggest that combining it with teacher support, practice with classmates, feedback, and pronunciation practice leads to more effective learning.

7. Educational Implications
This study indicates that when introducing AI in Chinese language education, simply letting students use it freely is insufficient. Teachers need to provide guidance on how to use AI itself.

Specifically, they point out the need to teach students not to blindly trust AI answers, to check if they match their own level, to change overly difficult expressions into simpler ones, and to ensure they are sentences they can pronounce themselves.

Also, while AI is useful for script writing and composition, they argue that it is important to link it to actual speaking practice to improve speaking skills. They suggest that incorporating pronunciation practice, pair work, teacher feedback, and impromptu response practice after using AI leads to more practical skills.

8. Research Limitations and Future Challenges
The authors list several limitations. First, the study is limited to one university in Malaysia, so the results cannot be directly applied to other countries or educational environments. Second, since the subjects are beginner learners, whether the same results apply to intermediate and advanced learners needs further investigation. Third, since Chinese proficiency before the start of the class was not sufficiently measured, the original ability differences among learners were not fully captured.

Future research is needed that targets different countries, regions, Chinese proficiency levels, and AI tools. It is also important to examine more detailed learning processes, such as how many times learners actually used AI, how they asked questions, and how they revised AI answers.

9. Reason for Choosing This Paper
The reason I chose this paper is that the theme of AI-supported Chinese speaking learning is deeply related to my own educational practice and research interests.

This paper is helpful because it analyzes AI not just as a convenient tool, but as something related to learners’ perceptions, attitudes, and learning behaviors. In particular, it provides many insights into how to use AI to lead to improvements in speaking skills and what kind of usage patterns become problematic.

AI is ultimately a tool to support learning, and the teacher’s role remains important. In that sense, this paper is a very useful study for thinking about Chinese speaking education in the AI era.

Written by: Na Li