Chapter One  Related Series

Chapter One mainly explains the new possibilities brought by AI and the “Double Reduction” policy to the teaching of spoken English in primary schools. At the beginning, let’s clarify the core concepts of this study, that is, the “Double Reduction” policy, AI-speaking coaches and oral English teaching, so that the boundaries of the research will be clearer. Then, this chapter sorts out the value of this work at the theoretical and practical levels, which paves the way for the discussion of the following chapters.

1.1 Research Background

Smartphones, social media and artificial intelligence (AI) are developing rapidly. For education practitioners, this rhythm brings both great challenges and new opportunities: they can use these tools to build more advanced learning media. In the past few decades, AI has been widely used in the development of various applications, and its products have penetrated into almost every corner of daily life.[1] This momentum has not stopped until now. Large-scale next-generation AI models represented by ChatGPT, Sora and DeepSeek have appeared one after another, driving a new round of AI boom and quickly spreading around the world. Zhang (2025) pointed out that generative AI has great application potential in the field of education because of its strong language understanding and personalized feedback ability.[2] Some academic insiders can’t help sighing that global education is now undergoing an unprecedented large-scale social practice trial.[3]

Modern technology continues to progress, and the global demand for practical English communication ability is also rising synchronously. These two trends are mutually driven. According to the data of Duolingo’s 2024 global report on language learning trends, the popularity of English is still at the top, covering 135 economies including China, and the number of learners has increased by more than 10% compared with the previous year.[4] There is a high degree of consensus in the academic community on this. Linguist Michael McCarthy believes: “Spoken English, as the primary form of everyday communication, is irreplaceable.”[5] David Crystal pointed out: “Spoken language is at the heart of all language use; to neglect research into spoken language is to neglect the very essence of language.”[6] English education expert Penny Ur also advocates: “Oral training should run throughout language teaching; Real-time verbal interaction in the classroom can slowly enhance students’ willingness to communicate confidently, and at the same time, solidify their base for reading, writing and other key language competencies.”[7]

The comprehensive implementation of China’s “Double Reduction” policy has brought a different opportunity to the transformation of oral English teaching in primary schools, and also pushed schools to constantly adjust the curriculum and make full use of existing resources.[8]

But if you really walk into the classroom and observe, you will find that the gap between the ideal setting and the actual teaching is quite obvious. Because teaching has been focused on written exams for a long time, exercises such as mechanical copying still account for a considerable proportion in classroom activities. This results in the frequent marginalization of speaking tasks. Especially in large-class teaching scenarios, it is difficult for teachers to provide timely and accurate pronunciation correction for each student. Class hours are limited, and students have few individual expression opportunities. As time goes on, the children’s willingness to speak is slowly worn away. The lack of targeted guidance ultimately leads to the accumulation of pronunciation difficulties. Pupils then fall into the inescapable dilemma of “silent English”.[9] 

To solve these practical difficulties, it is urgent to build a diversified monitoring system. This system should take in process-oriented evaluation and multi-dimensional feedback, and really fill in the gap between “textual standards” and “classroom practice” in primary school English teaching.[10]

1.2 Definition of Related Important Concepts

Firstly, the “Double Reduction” policy refers to the ‘Opinions on Further Reducing the Burden of Homework and Off-campus Tutoring for Students in Compulsory Education’, issued in July 2021 by the General Office of the Central Committee of the Communist Party of China and the General Office of the State Council, and commonly abbreviated as “Double Reduction”. The “Double Reduction” policy issues a call. Education quality must rise significantly. Teaching quality must rise significantly. Pupils must learn sufficiently at school. They must learn effectively at school. The policy makes an explicit emphasis. Classroom teaching quality needs a boost. Schools must improve teaching management procedures. Teaching methods need optimization. This part of teaching administration should be grasped. The efficiency of students’ study in school should be effectively improved, and it is stipulated that none of the content to be taught must be missing, so as to ensure that students meet the national academic quality standards. The class schedule should not be increased or decreased at will, the difficulty should not be increased arbitrarily, and the teaching progress should not be deliberately accelerated. The pressure on students in the examination should be gradually reduced, and the evaluation method should also be changed in a more reasonable direction. It is forbidden to end the new course early in order to prepare for the exam. Standardized examinations organized in violation of regulations are resolutely not to be done, and the examination questions must be controlled within the scope of the curriculum standards. The practice of queueing and labelling students according to the exam results should not be allowed; the results should be presented using a grading system, and the tendency of only looking at scores and taking scores as the only standard must be reversed with determination.[11]

Secondly, the word “AI” was officially named at the Dartmouth Conference in the United States in 1956.[12] Today, AI has spread to a lot of fields such as medical care, education, finance, industry and transportation. This study focusses on what role AI can play in education. “AI Speaking Coach” is a very specific example of how AI can be used in language education. It refers to a kind of application or system. To build this kind of oral coach, the underlying technology relies on large educational language models, which integrate intelligent speech recognition, natural language processing and adaptive learning algorithms into one to form a system that can understand speech, process language, and adjust according to the learner’s situation. The main goal of this set of things is naturally to practice students’ oral skills. The path of realization is to simulate real dialogue and give feedback immediately. Students practice oral communication through these simulated dialogues, and they are more willing to participate in the learning process as they practice. In this study, “AI Speaking Coach” does not refer to a specific App, but also a general term for digital tools such as “Fluent Talk”, “Speaking 100” and “Fun Dubbing”. Its most core value lies in the “one-to-one” virtual tutoring environment, where each student’s oral practice can get personalized feedback immediately. Relying on this kind of feedback, students’ oral communication skills will gradually improve, and their confidence in oral expression will gradually become more satisfied.

Third, there are many levels of oral English teaching itself, and each dimension is closely linked by the inherent logical relationship. Scholar Zeng Xianzhao mentioned that oral teaching is a core driving force, which can help primary and secondary school students transform language knowledge into practical skills. Through oral practice, students absorb fragmentary knowledge points such as pronunciation, grammar and vocabulary little by little, and then integrate them into communication skills that can be expressed naturally. At the same time, output-oriented training has also promoted the organic integration and coordination of listening, speaking, reading and writing skills.[13] This study believes that oral English teaching is actually a process of meaning construction of the target language with English, and the underlying basis is the Constructivist Learning Theory. The theory argues that learning is the process in which learners take the initiative to construct knowledge. Specifically, each learner relies on personal cognition to construct meaning, and also to construct meaning through social interaction. The real sense of learning lies in meaning construction itself, and there is no other alternative path.[14] Looking down from this theoretical perspective, the focus of classroom teaching is no longer on teachers’ teaching work, but focusing more on students’ independent learning. Correspondingly, the role of teachers has also changed, becoming the supporters and motivators of students throughout the learning process.[15] Stephen Krashen’s second language acquisition theory provides good psychological support for the application of AI in the field of education. According to the Input Hypothesis in this theory, the premise of effective language acquisition is that the learner must receive comprehensible input. The most ideal learning materials should be only a little higher than the students’ current language level, which is the so-called i+1 learning principle.[16] In practical research, teachers can instantly generate and provide language input adapted to each student with the help of “AI Speaking Coach”. Doing this just bypasses a common problem in traditional large-class teaching, the teaching arrangement is one-size-fits-all, and everyone rushes forward with a set of things.

1.3 Research Significance

This study has real and practical value, as shown below:

To begin with, the AI Speaking Coach benefits students. It creates a non-stop online interactive learning environment for primary school children. Under the “Double Reduction” policy, children’s after-school homework is less, and they can use this AI platform for short-term and high-frequency oral practice. This can not only improve pronunciation accuracy and speaking fluency, but also will not bring additional mental pressure. The system gives instant feedback, and students can detect and correct pronunciation errors in time. Slowly, speaking confidence will be established, and anxiety about speaking English will naturally be reduced.

In addition, this kind of smart tools can also help teachers allocate teaching resources more reasonably. In large-size classes, it is difficult for teachers to take care of the oral practice effect of each student, and the quality of the practice cannot be guaranteed. “AI Speaking Coach” can take on repetitive work, such as repeated pronunciation drills and simple dialogue exercises. In this way, teachers can devote more energy to creative lesson planning, emotional connection with students, and the cultivation of cross-cultural communication skills.

Not only that, the behavioral and learning data accumulated by AI can also help teachers evaluate the progress of each student more fairly. With this foundation, it becomes much easier to implement individualized and differentiated teaching.

Chapter Two  Current Situation and Problem Analysis

This chapter focusses on the difficulties faced by the teaching of spoken English in primary schools. Through the questionnaire survey of students and teachers separately, it was found that two problems were more prominent: first, the pronunciation guidance did not keep up well, and the other was that students’ oral anxiety was very common in the case of large classes. These results provide empirical support for the action research of “AI Speaking Coach” in Chapter 3.

2.1 Survey on English oral Learning for Primary Students

In order to understand the actual situation of oral English learning from the students’ side, this study specially designed a questionnaire for Class 2 of Grade 4. There are 15 questions in the questionnaire, mainly focusing on the following aspects:

Dimension 1: Interest in learning spoken English.

Dimension 2: Independent learning ability.

Dimension 3: Acceptance of AI.

A total of 55 questionnaires were distributed in this survey, of which 50 valid responses were recovered, with a response rate of 90.91%. The specific division of dimensions can be seen in the following table.

Table 2.1.1 Questionnaire division

Dimension	Question number
Students Interest	T1-T5
Self-regulation	T6-T10
Acceptance	T11-T15

Based on the scoring of the response options (A–D scored 4–3–2–1), the average score of each dimension is divided into three levels: Grade A (Good, 3–4 points), Grade B (Fair, 2–3 points) and Grade C (Pass, 1–2 points).

Table 2.1.2 Survey Questionnaire Dimension Level Calculation Table

Dimension	Grade A	Grade B	Grade C	Total
Students Interest	29.2	40.0	30.8	100
Self-regulation	16.7	54.1	29.2	100
Acceptance	19.3	48.8	31.9	100

Table 2.1.3 Interest in Oral English Learning

Question	Option	Frequency	Percentage
1.Do you like speaking English with others?	A. Like very much	8	16.0
	B. Like somewhat	15	30.0
	C. Neutral	14	28.0
	D. Dislike	13	26.0
2. Do you like using AI software for dubbing?	A. Like very much	18	36.0
	B. Like somewhat	12	24.0
	C. Neutral	12	24.0
	D. Dislike	8	16.0
3. Do you find oral English classes interesting?	A. Very interesting	16	32.0
	B. Quite interesting	19	38.0
	C. Neutral	10	20.0
	D. Boring	5	10.0
4. Do you want to speak English outside of class?	A. Often	9	18.0
	B. Sometimes	11	22.0
	C. Occasionally	18	36.0
	D. Not at all	12	24.0
5. Do you think speaking English is cool?	A. Very cool	20	40.0
	B. Quite cool	16	32.0
	C. Neutral	9	18.0
	D. Indifferent	5	10.0

The survey results indicate that whilst many students consider speaking English to be ‘cool’, only 46% actually enjoy communicating with others in English. This shows that students’ perception of spoken English is merely superficial. It has not turned into a strong desire to actively participate in oral practice. At the same time, students are highly dependent on interesting learning tools. They prefer “AI voice-overs” far more than traditional “classroom speaking classes.” Without the support of such engaging tools, they find it hard to build intrinsic learning motivation. Their initiative to do extracurricular oral practice is also very weak. Only a small number of students “often want to” speak English. Those who “occasionally want to” or “have no desire to” account for 60% together. Spoken English learning is severely disconnected from daily life, and students have failed to develop good language usage habits.

Table 2.1.4 Students’ ability to learn independently

Question	Option	Frequency		Percentage
6. Do you actively raise your hand to answer questions?	A. Always		5	10.0
	B. Often		11	22.0
	C. Occasionally		28	56.0
	D. Rarely		6	12.0
7.Do you look for software to practice on your own after class?	A. Always		6	12.0
	B. Often		9	18.0
	C. Occasionally		20	40.0
	D. Never		15	30.0
8. Do you do your oral English homework carefully?	A. Very carefully		21	42.0
	B.Quite carefully		19	38.0
	C. Average		7	14.0
	D. Not carefully		3	6.0
9. Do you look up words you don’t know how to pronounce by yourself?	A. Always		7	14.0
	B. Often		13	26.0
	C. Occasionally		21	42.0
	D. Never		9	18.0
10. Can you stick to practicing for a few minutes every day?	A. Absolutely		6	12.0
	B. Mostly		8	16.0
	C. Occasionally		19	38.0
	D. Cannot		17	34.0

Only 10% of the students who can regularly volunteer to join in oral activities in the classroom, and the overall participation enthusiasm is low. Most students have been in a passive state, waiting to be nominated by the teacher, and rarely raise their hands by themselves. Looking at the after-class practice habits, only 12% of those who insist on doing oral exercises every day, while 30% of the students who never take the initiative to self-practice with learning apps. Without ongoing supervision from the teacher, it is difficult for most children to sustain regular practice. Furthermore, students’ learning behaviors is highly dependent on external instructions. Although they are able to cooperate well in completing the oral assignments set by the teacher, the proportion of students who proactively look up unfamiliar words when they encounter them is extremely low. Their learning is largely driven by a desire to ‘get the task done’, and they have not mastered scientific methods of independent learning.

Table 2.1.5 Student Acceptance of AI Oral Tools

Question	Option	Frequency	Percentage
11. Do you hope to use AI oral software to help you learn?	A. Strongly hope	43	86.0
	B. Somewhat hope	5	10.0
	C. Indifferent	1	2.0
	D. Do not need	1	2.0
12. Is speaking to AI software more relaxing than speaking to a teacher?	A. Very relaxed	36	72.0
	B. Somewhat relaxed	9	18.0
	C. Same	4	8.0
	D. Nervous	1	2.0
13. Do you like the timely feedback from AI?	A. Like very much	33	66.0
	B. Like somewhat	10	20.0
	C. Neutral	5	10.0
	D. Dislike	2	4.0
14. Are you afraid of making mistakes in front of AI?	A. Never afraid	30	60.0
	B. Occasionally afraid	12	24.0
	C. Often afraid	5	10.0
	D. Always afraid	3	6.0
15. Do you usually use AI oral apps in your daily practice?	A. Always	5	10.0
	B. Often	8	16.0
	C. Occasionally	22	44.0
	D. Never	15	30.0

Survey data shows that AI speaking tools can provide students with a virtual practice environment with lower anxiety, timely feedback and greater acceptance of oral errors. Around 72% of participants said they felt much less nervous chatting with AI than talking to their teachers in person. In traditional large classes, immediate on-the-spot corrections and being watched by classmates easily generate mental stress.

Therefore, students clearly need a personalized, low-pressure learning space. They can practice freely and make mistakes without worry there. Some 60% of learners feel no nervousness when using AI platforms. Students generally regard the error prompts given by AI as useful suggestions for improvement, rather than criticizing or picking mistakes. The instant feedback function is also very popular, which just makes up for the long waiting time in traditional homework evaluation, and the students’ enthusiasm was consumed.

However, on the other hand, students’ attitude towards AI learning tools is quite positive, but after class, there are not many people who can take the initiative to pick up and use them. The recognition rate is very high, but the actual use rate is low, and there is a clear gap in the middle. This also more or less shows one thing: the traditional large-class instruction is actually difficult to take care of the individual guidance needs of each student. Therefore, it is necessary to rub AI resources into the daily teaching design and take students to form long-term and stable oral practice habits step by step.

2.2 Survey on English oral Learning for Primary Teachers

In order to better understand the current situation of oral English teaching on the teacher’s side, the researchers made another questionnaire survey. The sampling principle is consistent with the student survey, and the survey covers all the English teachers in the fourth grade of Xiadongying Primary School. There are 10 questions in the questionnaire, mainly asking about the bottlenecks encountered in teaching at ordinary times and their views on the use of AI educational tools. The questionnaire was sent to five fourth-grade English teachers, and all five were collected, and all answers were valid, with a recovery rate of 100%.

Table 2.2.1 Survey Results of Grade 4 English Teachers (N=5)

Question	Option	Frequency	Percentage
1.What is the average class size you teach?	A. 30 and below	0	0.0
	B. 31-40	0	0.0
	C. 41-50	3	60.0
	D. Above 50	2	40.0
2. What is the frequency of dedicated oral practice you can arrange per week?	A. Every class has sufficient time	0	0.0
	B. 1-2 times a week	1	20.0
	C. Occasionally	3	60.0
	D. Rarely	1	20.0
3. Can you correct every student’s pronunciation?	A. Completely possible	0	0.0
	B. Mostly possible	0	0.0
	C. Occasionally possible	2	40.0
	D. Hardly possible	3	60.0
4. Do students show speaking anxiety in class?	A. Rarely	0	0.0
	B. Occasionally	1	20.0
	C. Quite common	3	60.0
	D. Very severe	1	20.0
5. How heavy is the burden of grading students’ oral homework, such as listening to audio recordings?	A. Very relaxed	0	0.0
	B. Moderate burden	0	0.0
	C. Heavy burden	1	20.0
	D. Extremely heavy burden	4	80.0
6.How familiar are you with AI Speaking coaches?	A. Very familiar	0	0.0
	B. Somewhat familiar	1	20.0
	C. Just heard of them	3	60.0
	D. Completely unfamiliar	1	20.0
7. Do you support the introduction of intelligent tools like “AI Speaking Coach” to share basic pronunciation correction tasks?	A. Strongly support	4	80.0
	B. Somewhat support	1	20.0
	C. Neutral	0	0.0
	D. Do not support	0	0.0
8. Do you agree that AI’s environment can ease student anxiety?	A. Strongly agree	3	60.0
	B. Somewhat agree	2	40.0
	C. Unsure	0	0.0
	D. Disagree	0	0.0
9. How do you view the value of AI-generated visual learning reports for your daily teaching?	A. Huge help	3	60.0
	B. Some help	2	40.0
	C. Little help	0	0.0
	D. No help	0	0.0
10.  Are you willing to integrate AI Speaking tools into your regular teaching practices?	A. Very willing	4	80.0
	B. Somewhat willing	1	20.0
	C. Depends	0	0.0
	D. Not willing	0	0.0

Looking at the results of the questionnaire together, several key problems encountered in the traditional fourth-grade oral English teaching will be clear. In the classroom, the teacher can’t give enough care to every student, and the burden of correcting homework after class is particularly heavy. Sixty presents of the interviewed teachers admitted that they could hardly provide personalized guidance and instant pronunciation correction to each child. The traditional assessment also consumes a lot of time and energy for teachers. As many as 80% of teachers find it quite tiring to correct after-class oral homework, and it is naturally difficult to give timely feedback to students. In addition, 80% of teachers (C and D options totaled) pointed out that fear of making errors is a common psychological hurdle among students, which is in match with the results of the student survey in Section 2.1.

In terms of teachers’ acceptance of digital teaching innovation, the contrast in the data is quite obvious: they have little actual contact, but they are full of enthusiasm to try. 60% of teachers only stick to the shallow cognitive level of AI teaching tools. They have not really used them at hand, but they are eager to change the existing teaching model. Up to 80% of teachers fully support the introduction of AI speaking coaches to do routine work such as basic phonetic practice. At the same time, all participating teachers (100%, options A and B are added) agree that the low-pressure, non-critical interactive atmosphere of AI can really relieve the nervousness of students when they speak English in public.

Chapter Three  Implementation Based on AI Speaking Coach

The research conclusions in Chapter Two can be summed up mainly in the following aspects: classroom speaking anxiety, insufficient self-learning capacity and teachers’ heavy workload. Drawing from these, the author combined the “Double Reduction” policy with the general background of educational digital reform, and carried out four-month action research. The research subjects were 55 fourth-grade students. With the help of large-model-based AI speaking assistance, a teaching framework covering pre-class learning preparation, in-class oral expression and post-class knowledge absorption was built.

3.1 Design and Preparation of Action Research

At the beginning of the research, the preliminary survey found that students and teachers had their own difficulties. As for students, there are few students who can take the initiative to communicate in English in traditional classes. Most of them simply don’t make a sound because they are afraid of making mistakes and nervous. As for the teacher, under the large class quota, there is one-on-one pronunciation correction for each student. I really can’t take care of it. After class, I have to correct a bunch of after-class oral assignments, and I can’t finish it at all. For these two problems, this study believes that “AI Speaking Coach” is a relatively ideal solution. Its set of multimodal resources can provide high-quality original voice demonstration and very life-like dialogue scenes. Students can directly imitate and slowly internalize and absorb them. In addition, the automated assessment feature also helps to reduce the burden of correction, ensuring that each student can get their own personalized, real-time feedback.

The practical content of this research covers the six thematic units of the PEP Grade 4 Volume 1 English textbook. The overall learning goal is as follows: with the help of AI, students should be able to accurately recognize and read the core vocabulary in these different daily topics, and can also apply the core sentence patterns more skillfully to complete meaningful communication expressions in the corresponding real situations.

3.2 Implementation Plans for Teaching Practice

In the actual implementation of primary school English speaking instruction, this study intends to focus not only on a single lesson case, but includes all six units of the PEP Grade 4 Volume 1 textbook. Dealing with it in this way directly bypasses the narrowing limitation of just watching a lesson. Several core functional modules of the AI speaking coach are closely embedded in three teaching phases: pre-class independent perception, in-class interactive output and post-class precise internalization. In practical operation, two units will be selected as demonstration cases at each teaching phase, directly aimed at the most difficult points in traditional oral English teaching.

Table 3.2.1 Teaching Plan

Stages	Student Application	Teacher Application	Theme
Pre- class	With AI support, students recognize the vocabulary related to the classroom and schoolbag, and follow the 3D lip-shape animations on the screen to read aloud. After reading, they can immediately get individual scoring and targeted pronunciation correction.	Teacher arranges pre-learning vocabulary assignments through the AI platform, and then looks at the backend learning data, and it is easy to find the shared pronunciation problems of the students in the class. In this way, the gap that personalized guidance can’t keep up in large-class teaching can be filled.	Unit 1 My classroom Unit 2 My schoolbag
While-class	The low-pressure, bias-free conversations between students and AI is carried out. The tasks completed include describing looks, searching for items and other activities. In this way, their fear of speaking English will gradually be reduced.	In actual teaching, we design a vivid AI-supported role-play contexts, no longer relying on rote memorization, but promoting practice with the help of interactive tasks with emotional encouragement.	Unit 3 My friends Unit 4 My home
Post-class	Students first complete scenario dubbing and some daily communication tasks, such as ordering food and introducing family members, and then use AI for self-summary and knowledge consolidation on the AI’s judgement on intonation and oral fluency.	Assign the after-class tasks of smart voice recording and scenario dubbing assignments, and directly use visual data reports to carry out targeted instruction. This effectively lightens the huge load of traditional manual assessment.	Unit 5 Dinner’s ready Unit 6 Meet my family

Pre-class Independent Perception (Units 1 & 2): Tackling Pronunciation Difficulties and Lack of Personalized Instruction in Large Classes. In large class teaching, teachers cannot correct the pronunciation of each student one by one, so they rely on AI technology to fill this gap. In the pre-class preparation of Unit 1 My classroom and Unit 2 My schoolbag, students independently learn vocabulary related to classroom facilities and school supplies, read and practice pronunciation through AI tools. The platform gives 3D lip animation and instant scoring, so that each student can get tailored basic pronunciation guidance before class, instead of eating big pot rice.

While-class Interactive Output (Units 3 & 4): Constructing Real-life Scenarios to Ease Speaking Anxiety. In response to the old problems of thin classroom situation and students’ fear of making mistakes, AI-created immersive role-play scenes are used in the teaching of Unit 3 My friends and Unit 4 My home. The scene is designed around practical communication tasks such as describing appearance characteristics and finding daily necessities. Students don’t have any burdens. They don’t have to be afraid of being judged. They talk to AI very relaxedly, like “Where are the keys?” You can practice it just by saying it casually. The low-pressure learning atmosphere makes everyone’s oral learning tension obviously disappear a lot, and the vigor of raising their hands to participate in class has suddenly increased a lot.

Post-class precise internalization (Units 5 & 6): Smart Auto Evaluation to Reduce Teachers’ Assessment Workload. In order to relieve the heavy burden of after-class oral homework correction, after teaching Unit 5 Dinner’s ready and Unit 6 Meet my family, we assigned AI-enabled voice recording and scenario dubbing tasks. The intelligent system can judge the intonation and fluency of students by itself, and can also generate visual progress reports immediately. As for the students, the relevant sentence patterns of ordering food and introducing family members are written down steadily over and over again, and they are more lively than before. On the teacher’s side, they saved a lot of time listening to recordings and scoring, which laid the way for data-supported personalized teaching in the future.

3.3 Effectiveness and Reflection

Using the same marking scheme (pronunciation 40% + fluency 30% + content 30%), pre- and post-tests were conducted on 55 students. The results are as follows:

Table 3.3.1  Oral Test Score Comparison（N=55）

Grade (Score)	Pre-test Participant Count	Percentage (%)	Post-test Participant Count	Percentage (%)	Variation Trend
Excellent (85+)	5	9.1	16	29.1	↑20.0%
Good (70-84)	18	32.7	25	45.5	↑12.8%
Passing (60-69)	22	40.0	12	21.8	↓18.2%
Needs Improvement (<60)	10	18.2	2	3.6	↓14.6%
Average Score	68.4		78.9		↑10.5

The class average in oral performance increased from 68.4 in the pre-test to 78.9 in the post-test, which is a real improvement. It can be seen from this increase that the use of AI resources is indeed of substantial help to students’ oral ability. Looking at the segmentation, the excellent rate jumped from 9.1% to 29.1% at once, which shows that instant feedback and targeted pronunciation support of the AI system have helped students break through some stuck places in oral expression. The ratio of failing students has also dropped significantly, and the number of learners needing remediation has been reduced from 10 to 2. This result can at least explain two points: the stress-free practice environment built by AI can greatly ease the speaking anxiety of later students when speaking English, and at the same time help them fix learning deficiencies. In the end, the significant increase in test scores supports the basic judgement of this study - under the framework of the Double Reduction policy, AI tools are embedded in daily oral English teaching, students’ phonetic intonation can be optimized, and a very practical path can be found for the improvement of the quality of primary school English teaching and classroom efficiency.

After more than four months of practice, we can indeed see some real changes in the classroom. From the perspective of students, AI is more like a very patient companion. Children who are usually afraid to open their mouths because they are shy and afraid of making mistakes were also brought in, and everyone’s overall tension was obviously much lighter. Of course, there was also a little situation when I first started: in order to get a higher AI rating, several students deliberately said it quickly, and the sentences jumped out stiffly, as if they were memorizing books, without real emotions at all. As soon as the teacher found this, he immediately adjusted the evaluation criteria, requiring students not only to achieve precise pronunciation, but also to take expressive intonation seriously. With this change, it avoids mechanical rote reading, and slowly leads the child to a state of communication that is more like a real-life daily conversation.

From the perspective of teachers, AI solves the most troublesome problem in large class teaching. In the past, teachers had to spend a lot of time listening to the recording and correcting the pronunciation one by one. Now this part of the burden is much lighter. It was tiring, and students couldn’t get immediate feedback. Now, the AI does this job, and teachers no longer have to act as “correction machines.” With the time saved, teachers can focus on designing more interesting class activities or helping students who are falling behind. This perfectly matches the goal of the “Double Reduction” policy: reducing the burden while improving teaching quality.

Conclusion

This paper conducted four months of systematic and standardized action research at Xiadongying Primary School. The study confirmed that AI Speaking Coach is indeed useable in primary school English teaching, and can also promote meaningful educational innovation. The data shows that the classroom language atmosphere has changed significantly. The instant, objective and high-frequency feedback given by this intelligent tool is equivalent to adding a psychological buffer to students. Speaking practice no longer has to bear the strong interpersonal pressure of traditional classrooms, and the learning atmosphere becomes relaxed and safe. In such an atmosphere, students can try pronunciation and expression without fear of making mistakes. The common emotional barrier of young students in large-class settings has been weakened. Slowly, the students’ fear of opening up has subsided a lot, and the learning attitude has changed from the previous passive acceptance to the active want to participate, so that they can move forward.

Relying on the AI’s professional automatic pronunciation assessment system, coupled with immersive scene simulation, students’ pronunciation accuracy and speaking fluency are gradually improved. The help brought by technology has contributed to a very important cognitive shift, allowing learners to cross simple mechanical memorization and gradually establish more independent and natural oral expression. Timely on-site feedback can immediately correct inappropriate places and avoid wrong language habits from taking root in time, which also fills a big gap that has always existed in traditional classrooms, that is, teachers can hardly give one-on-one personalized guidance to each student.

After the introduction of AI, the division of teaching tasks has indeed become much more reasonable. The repetitive pronunciation training and the hard work of marking personalized oral assignments are all left to the intelligent system, so that teachers no longer have to waste on mechanical and endless error correction. In this way, teaching burnout can be really alleviated, and the professional autonomy in the hands of the teacher can also be regained. Now, they can spend more time on innovative curriculum design, giving emotional counseling to students, and targeted tutoring for underachieving students. These things are more humane and can better reflect the temperature of education.

Under the big plate of the national “Double Reduction” policy, introducing AI into teaching is no longer as simple as adding an auxiliary tool. It is actually a necessary action to achieve burden reduction and quality improvement. In order to maintain the teaching effect for a long time, this study proposes a complete closed-loop teaching mode, covering pre-class preparation, in-class interaction and post-class consolidation. This study also noticed that the digital divide is actually very obvious. Different home equipment and network conditions are likely to lead to unfairness in learning results, which needs to be properly paid attention to in practice.

In the end, this study is just a local exploration, with small samples and short research cycles. In addition, the self-reported questionnaire data may be biased, and students may report their interest to the teacher in order to cater to the teacher’s expectations. Whether the momentary freshness brought by the new technology can be maintained can only be judged by long-term tracking. For follow-up research, you may want to lengthen the time span and use more objective materials, such as speech analysis records, learning behavior tracks, etc., to use these methods to more accurately understand the long-term role of AI in language learning. Generally speaking, this study fully shows that in the wave of educational digital transformation, AI tools are indeed a key way to cultivate primary school students’ courage to express themselves and overall English literacy.

Bibliography

1. Min M M. AI Technology: A New Tool for Personalized Teaching in Primary School English Classrooms[J]. Exploration of Educational Management,2025,3(6):157-162.
2. 张海鹏,左一冉.生成式人工智能教育应用风险治理路径探索——基于TOE框架的系统分析[J].教育评论,2025,(08):95-104.
3. Nuria H. Artificial Intelligence (AI) Chatbot as Language Learning Medium: An inquiry[J]. Journal of Physics: Conference Series,2019,1387:012020.
4. Cindy Blanco, Ph.D. 2024 duolingo language report[EB/OL]. [2024-12-1]. https://blog.duolingo.com/2024-duolingo-language-report/.
5. MC CARTHY M. Spoken Language and Applied Linguistics[M]. Cambridge: Cambridge University Press,1998:1.
6. CRYSTAL D. The Cambridge Encyclopedia of the English Language[M]. Cambridge: Cambridge University Press,2018:401.
7. UR P. A Course in Language Teaching: Practice and Theory[M]. Cambridge: Cambridge University Press,2012:120.
8. 潘昕怡.如何实现“双减”政策背景下小学英语教学的提质增效[J].教育,2025,(30):49-51.
9. 张明瑶.移动软件“少儿趣配音”在小学英语口语教学中的应用研究[D].兰州:西北师范大学,2023.
10. 周诗杰.学业质量标准：从“文本”到“课堂”的实践困境与跨越——基于《义务教育英语课程标准（2022年版）》的学理分析与操作进路[J].小学教学研究,2025,(36):12-16.
11. 黄建云.“双减”背景下提升小学英语教学质量的管理策略研究[D].大理:大理大学,2024.
12. McCarthy J, Minsky M L, Rochester N, et al. A proposal for the dartmouth summer research project on artificial intelligence, august31,1955[J]. AI magazine,2006,27(4):12-12.
13. 曾宪灶.山区县中小学英语口语教学困境与人工智能的破局路径探究[J].校园英语,2026,(02):195-198.
14. Hein G E. Constructivist learning theory[J]. Institute for Inquiry,1991,14.
15. Bada S O, Olusegun S. Constructivism learning theory: A paradigm for teaching and learning[J]. Journal of Research & Method in Education,2015,5(6):66-70.
16. KRASHEN S D. The input hypothesis：issues and implications[M]. New YorkLongman,1985:2.

      

Acknowledgments

My deepest gratitude goes first and foremost to Ms. Zou, my tutor, for her painstaking reading of this thesis, valuable suggestions, and unwavering patience in helping me accomplish this final draft. Her profound academic knowledge and rigorous research attitude have not only guided me through the complexities of this paper but also set a brilliant example for my future career.

I would like to express my heartfelt appreciation to my family and friends, who have been my strongest pillars of support throughout this journey. I am deeply indebted to my parents for their unconditional love and selfless sacrifices; their constant encouragement provided me with a safe harbor and the courage to pursue my academic goals. My sincere thanks also go to my dear friends and classmates. Their companionship during late-night study sessions and their willingness to listen to my concerns have been a powerful "uphill push," turning many moments of doubt into motivation and inspiration. The vibrant atmosphere and mutual support we shared have made these years truly unforgettable.

Finally, I would like to extend my sincere gratitude to all the teachers and professors in the English Department for their dedicated teaching and help during my four years of college life. Their expertise and guidance have not only broadened my academic horizons but also prepared me for the challenges of the future.

Appendix A Questionnaire in Chinese

小学四年级英语口语学习现状调查问卷（学生版）

第一维度：口语学习兴趣（5题）

( ) 1. 你喜欢上英语课跟老师或同学一起说英语吗？

A.非常喜欢

B.比较喜欢  

C.一般

D.不喜欢

( ) 2. 你喜欢跟着英语歌、动画片或配音APP一起读英语吗？

A.非常喜欢  

B.比较喜欢  

C.一般  

D.不喜欢

( ) 3. 英语课上，老师让大家举手读课文或对话时，你会主动举手吗？

A.每次都举

B.经常举

C.偶尔举  

D.从不举手

( ) 4. 如果有英语角或者英语小剧场活动，你愿意参加吗？

A.非常愿意  

B.比较愿意  

C.看情况  

D.不愿意

( ) 5. 你喜欢在班里或者家人面前表演英语小对话吗？

A.非常喜欢  

B.比较喜欢  

C.一般  

D.不喜欢

第二维度：自主学习能力（5题）

( ) 6. 除了英语课作业，你会自己主动听英语、读英语或者用APP练口语吗？

A.经常主动

B.偶尔主动  

C.家长催才做  

D.从不

( ) 7. 遇到不会读的英语单词或句子，你会自己查词典、问家长或用APP听发音吗？

A.每次都会  

B.经常会

C.偶尔会  

D.不会，等老师教

( ) 8. 你会自己跟着课本录音或APP一句一句跟读英语吗？

A.经常

B.有时  

C.很少  

D.从不

( ) 9. 你会把自己读的英语录下来，听听哪里读得不对吗？

A.经常  

B.有时  

C.很少  

D.从不

( ) 10. 放假或周末在家，你会自己拿起英语书读一读吗？

A.经常  

B.有时  

C.很少  

D.从不

第三维度：对AI口语工具的接受度（5题）

( ) 11. 你用过能打分或纠正发音的英语口语APP（如一起作业、英语趣配音、伴鱼等）吗？

A.经常用  

B.偶尔用  

C.用过一两次

D.没用过

( ) 12. 如果用AI给你读英语的发音打分（比如85分），你愿意跟着练到更高分吗？

A.非常愿意  

B.比较愿意

C.一般  

D.不愿意

( ) 13. 你愿意让AI（而不是真人老师）来教你读单词、纠正你的发音吗？

A.非常愿意

B.比较愿意  

C.一般  

D.不愿意

( ) 14. 如果AI能像朋友一样和你用英语对话聊天（不是考试），你愿意尝试吗？

A.非常愿意  

B.比较愿意  

C.一般

D.不愿意

( ) 15. 比起对着课本读，你更喜欢用AI口语软件练发音吗？

A.更喜欢软件  

B.差不多  

C.更喜欢课本  

D.都不喜欢

小学四年级英语口语学习现状调查问卷（教师版）

1. 您目前任教的班级，平均学生人数约为？

A. 30人及以下

B. 31-40人

C. 41-50人

D. 50人以上

2. 受限于课时安排，您每周能为学生安排专门口语交际互动（如情景对话）的频率是？

A. 每节课都有充足时间

B. 每周1-2次

C. 偶尔安排

D. 几乎不安排

3. 在常规的大班教学中，您能兼顾并逐一纠正班级每位学生的口语发音吗？

A. 完全能兼顾

B. 基本能兼顾

C. 偶尔能兼顾

D. 很难兼顾

4. 在课堂口语互动环节，您观察到学生普遍存在开口焦虑或惧怕犯错的心理吗？

A. 几乎没有

B. 偶尔存在

C. 比较普遍

D. 非常严重

5. 在批改学生的口语作业（如听取语音记录）时，您的主观负担感受是？

A. 比较轻松

B. 负担适中

C. 负担较重

D. 负担极大

6. 您认为当前阻碍您提升学生口语水平的最大难点是？

A. 学生缺乏兴趣

B. 缺乏真实的交际语境

C. 班级人数过多，难以个性化指导

D. 缺乏科学客观的评价手段

7. 您是否支持引入“AI口语教练”等智能工具来分担基础发音纠正任务？

A. 非常支持

B. 比较支持

C. 持保留态度

D. 不支持

8. 您是否认同 AI 提供的“非评判性对话环境”能有效缓解学生的开口焦虑？

A. 高度认同

B. 基本认同

C. 不确定

D. 不认同

9. 您如何看待 AI 系统自动生成的“可视化学情报告”对教学的帮助？

A. 帮助巨大

B. 有一定帮助

C. 帮助较小

D. 几乎没有作用

10. 面向未来，您是否有意愿将 AI 口语工具常态化地纳入您的日常教学中？

A. 意愿非常强烈

B. 比较有意愿

C. 视情况而定

D. 暂无意愿

Appendix B Questionnaire in English

A Survey of Grade Four English Students on Oral Instruction

Dimension 1: Interest in Speaking (5 questions)

() 1. Do you enjoy speaking English with your teacher or classmates during English class?

A. Very much

B. Somewhat

C. Neither here nor there

D. Not at all

() 2. Do you enjoy reading along with English songs, cartoons, or voice-over apps?

A. Very much

B. Somewhat

C. Neither here nor there

D. Not at all

() 3. In English class, when the teacher asks students to raise their hands to read a passage or dialogue, do you volunteer?

A. I raise my hand every time

B. I raise my hand often

C. I raise my hand occasionally

D. I never raise my hand

() 4. If there were an English corner or a short English play activity, would you be willing to participate?

A. Very willing

B. Somewhat willing

C. It depends

D. No

() 5. Do you like performing short English dialogues in front of your class or family?

A. Very much

B. Somewhat

C. Neither here nor there

D. Not at all

Dimension 2: Self-directed Learning Ability (5 questions)

() 6. Aside from English homework, do you take the initiative to listen to English, read English, or practice speaking using apps?

A. Often on my own

B. Occasionally on my own

C. Only when my parents remind me

D. Never

() 7. When you encounter English words or sentence you don’t know how to read, do you look them up in a dictionary, ask your parents, or use an app to check the pronunciation?

A. Every time

B. Often

C. Occasionally

D. No, I wait for the teacher to teach it

() 8. Do you practice reading English sentence by sentence along with textbook recordings or apps?

A. Often

B. Sometimes

C. Rarely

D. Never

() 9. Do you record yourself reading English to listen for mistakes?

A. Often

B. Sometimes

C. Rarely

D. Never

() 10. During holidays or weekends at home, do you pick up an English book and read on your own?

A. Often

B. Sometimes

C. Rarely

D. Never

Dimension 3: Acceptance of AI Speaking Tools (5 questions)

() 11. Have you used English speaking apps that score or correct pronunciation (e.g., Yuqi Zuoye, English Fun Dubbing, Banyu, etc.)?

A. Often

B. Occasionally

C. Once or twice

D. Never

() 12. If an AI scores your English pronunciation (e.g., 85 points), would you be willing to practice until you achieve a higher score?

A. Very willing

B. Somewhat willing

C. Neutral

D. Unwilling

() 13. Would you be willing to have an AI (rather than a human teacher) teach you how to pronounce words and correct your pronunciation?

A. Very willing

B. Somewhat willing

C. Neutral

D. Not willing

() 14. If an AI could chat with you in English like a friend (not for a test), would you be willing to try it?

A. Very willing

B. Somewhat willing

C. Neither here nor there

D. Not willing

() 15. Compared to reading from a textbook, do you prefer practicing pronunciation with AI speech software?

A. Prefer the software

B. About the same

C. Prefer the textbook

D. Don’t like either

A Survey of Grade Four English Teachers on Oral Instruction

1. What is the average class size you teach?

A. 30 and below

B. 31-40

C. 41-50

D. Above 50

2. What is the frequency of dedicated oral practice you can arrange per week?

A. Every class has sufficient time

B. 1-2 times a week

C. Occasionally

D. Rarely

3. Can you provide individual attention and correct every student’s pronunciation in a large class?

A. Completely possible

B. Mostly possible

C. Occasionally possible

D. Hardly possible

4. Do you observe widespread speaking anxiety or fear of making mistakes among your students during class?

A. Rarely

B. Occasionally

C. Quite common

D. Very severe

5. How heavy is the burden of grading students’ oral homework, such as listening to audio recordings?

A. Very relaxed

B. Moderate burden

C. Heavy burden

D. Extremely heavy burden

6. What do you consider to be the biggest obstacle to improving students’ oral skills?

A. Students’ lack of interest

B. Lack of authentic communication context

C. Large class size, difficult to provide personalized guidance

D. Poor objective evaluation methods

7. Do you support the introduction of intelligent tools like "AI Speaking Coaches" to share basic pronunciation correction tasks?

A. Strongly supports

B. Somewhat support

C. Neutral

D. Do not support

8. Do you agree that the "non-judgmental dialogue environment" provided by AI can effectively alleviate students’ speaking anxiety?

A. Strongly agrees

B. Somewhat agrees

C. Unsure

D. Disagree

9. How do you view the value of AI-generated visual learning reports for your daily teaching?

A. Huge help

B. Some help

C. Little help

D. No help

10. Looking ahead, are you willing to seamlessly integrate AI oral tools into your regular teaching practices?

A. Very willing

B. Somewhat willing

C. Depends

D. Not willing