Large Language Model (LLM-ChatGPT) and Learner Autonomy: Teaching Present Simple Tense as a Model ()
1. Introduction
Learner autonomy was firstly coined by Henri Holec (1980). He defines it as “learner’s ability to take charge of his/her own learning”. With autonomy, learners are in charge of their own learning process. It is also called self-directed learning because it puts the learner in the center of attention in language learning education. It was firstly twined with foreign language studies. With time span it has spread to many fields of educational studies and with diffusion of technology, it has become more related to E-learning [1].
Generally, one of the main reasons of low language learning performance lies in the absence of learning autonomy, although people are autonomous learners by instinct and autonomy is part of their lived experiences, due to many external and environmental factors, the level of language learners’ autonomy does not rank to the required level [2]. Alrabai [3], in his study on Saudi EFL learners, explains that they got low performance in English language international tests due to many factors. The most and foremost one is the absence of learners’ autonomy in the Saudi educational system.
This research endeavors to narrow the divide in promoting learner autonomy through the utilization of LLC-ChatGPT. EFL learners undergo training in utilizing this application to learn the rule of the present simple tense with minimal teacher intervention. This study aims to investigate the impact of using LLM-ChatGPT to teach the learners the present simple tense autonomously via providing grammatical rules assistance, automated feedback, and chances for regular drillings compared to the traditional methods of teaching grammar (over reliance on teacher). It also aims to gauge the students’ perceptions and attitudes of using LLM-ChatGPT as a reinforcement approach to learner autonomy. Out of these goals, two questions will be answered: What is the effect of using LLM-ChatGPT on enhancing learners’ autonomy? Is it more effective than the teacher-centered approach? What is the learners’ attitude/perception of using LLM-ChatGPT as a tool for autonomous learning?
The results demonstrate the efficacy of this application in teaching the learners the rule of present simple tense which is similar to the methodology employed by teachers in teaching the rule of the same tense. The findings conclude that LLM-ChatGPT is an effective tool that enhances learning present simple tense autonomously. It strengthens the learners’ memory and makes them remember the structure of the present simple tense through giving them automated feedback, rules assistance, and chances for constant practice. It also shows increasing desire towards learner’s centeredness and low interest to teacher’s dominant method.
2. Literature Review
2.1. Autonomous Learners: Beliefs and Behaviors
According to Holec [4], language learners are either autonomous or non-autonomous. He uses this dichotomy view to enhance the learners transition from being non-autonomous to be autonomous by boosting their ability for self-management. Little [5] modified this view and concentrated on bringing language learners’ autonomy engaged elsewhere to the process of language learning. Unanimously, researchers agree on the idea that learners’ autonomy enhances and impacts positively language proficiency of learners. It incorporates learners in process of language teaching and learning through defining their own goals and creating their own learning opportunities. It also makes them aware of their strengths and weaknesses and enhances self-confidence [2] [6] [7] [18]. Beliefs and action are two integrated and interrelated terms. Beliefs about autonomy does not make sense unless it comes into practice [8]-[12]. Many studies reported that there is an intertwined relationship between language learners’ beliefs and their learning behaviors. i.e. language learners’ beliefs about autonomy significantly impact their autonomous behaviors. Therefore, a regular expose of learners to autonomous practices, refines and reconstructs their beliefs about autonomy [13]-[16]. Conversely, acquiring wrong beliefs about autonomy may impact negatively the improvement of learners’ autonomy [17] [18].
2.2. LLM ChatGPT
The emergence of the field of Artificial Intelligence (AI) within computer science has facilitated the development of applications capable of exhibiting intelligent behaviors akin to human cognition. Consequently, this advancement has fundamentally altered societal paradigms, influencing cognitive processes, occupational practices, and lifestyles. Owing to its efficacy in addressing complex challenges, AI is extensively deployed across diverse domains of practical application, including but not limited to medicine, finance, and engineering …etc. [19].
After the emergence of chat-GPT, LLM gained immense popularity. In 2020, a new LLM called GPT-3 was able to perform many tasks with only a few examples of simple instructions [17]. In 2022, a new LLM approach introducing human customization in GPT showed that smaller models with human feedback produced more human-like and reliable results than larger model [20].
In almost all educational fields, scholars and researchers actively investigate the potential uses of LLMs. For example, in media and journalism, there is an increasing recognition that LLMs is able to elevate the quality of work produced, so students should be familiar with these tools in order to utilize them effezctively. In medical education, chat-GPT helps students in the USA to score good grades in United States Medical Licensing Exam (USMLE) [21]. This event may positively impact student assessment methods and the course materials design [22]-[24].
Artificial intelligence language models (AI-LM) are currently used in achieving many tasks related to language learning such as machine translation, text summarization and text generation [9]. Transformer language Model (TLM) introduced by Radford et al. [23], whose performance in responding to some language tasks becomes a good addition in language learning. It helps EFL learners understanding many language tasks such as answering questions and natural language inference. In translation and summarization, there is an application called GPT-3 introduced by Brown [5]. It shows great performance in achieving the tasks related to translating text between different languages and summarizing long texts. In writing research papers, XL-Net is another model used by Yang [9] [22]. It is able to comprehend the meaning of a word in a context and able to consider the relationship between words in a sentence which leads to understand the meaning of the entire text.
Natural Language Processing (NLP) is really amazing application. It allows computer to interpret texts and generate speech. It also supports virtual assistant like Siri and Alex which has the function to comprehend and simultaneously respond to spoken commends [25].
These language learning models are used in a large amount of text data. By applying neural networks to learn and generate human-like text, they can perform various language tasks and trigger potential applications in different language learning fields. Therefore, IT applications progress rapidly in the learning languages area because of the dire need for advances in deep learning, and creative applications of algorithms to enhance language learning.
3. Ethical Consideration
In adherence to ethical considerations, the researcher adhered to the principles outlined in the Helsinki Declaration. Prior to commencing the study, he obtained a consent letter from the Chairman of the English Department, thereby ensuring procedural integrity and respect for the participants’ autonomy. Furthermore, the researchers obtained oral consent from the selected students who demonstrated willingness to participate in the study. These students were assured that their identities would remain confidential and that their scores would be utilized solely for research purposes, aligning with ethical standards of confidentiality and informed consent.
4. Methodology and Procedures
In light of the proliferation of technology and the burgeoning interest among the younger generation in its tools, the current study tends to use one of AI tools called Language Learning Model (LLM-ChatGPT) to enhance language learning autonomy. It will be adopted and incorporated in the process of learning Present Simple Tense. The total number of students who undergo the course (Grammar I) is 80 and are distributed into 4 sections. Out of this number, only 50 students would participate in this study. They are enrolled in first semester, Department of English, King Khalid University. Therefore, the sample of the study is selected randomly and each student was assigned to the one of the two groups in an unbiased selection. A between subject design using control group (N = 26) and experimental group (N = 24) is the methods used in this study. The experimental group (N = 24) will be trained on how to use LLM chat-GPT including opening personal accounts, being familiar with its functions, and using customization icon to give orders and get automated feedback about the structure of Present Simple Tense without teacher’s interference. LLM-ChatGPT software will inform students about grammatical mistakes they make in present simple tense without specifying them and provide two attempts for correction. Should the two initial attempts prove unsuccessful, GPT in the third iteration will provide the correct sentence, thereby serving as a remedial measure to enhance students’ retention about the structure of present simple tense. The control group (N = 26) will be taught the same tense in traditional method with the teacher as the center of the learning process. Pre-and post-tests will be conducted to know if there are significant differences between the two-group’s performance and to what extent LLM-ChatGPT enhances learner autonomy.
A Likert-scale questionnaire will be distributed on the students of the experimental group to know their attitude/perception about using LLM ChatGPT to enhance learner autonomy through applying this application on learning present simple tense.
The study has three sessions; in session one, all students in the two groups would be given pre-test to measure their performance in present simple tense. The pre-test shows no significant differences in the mean score of the control and experimental groups. They are almost in the same level of achievement. The second session would be allotted to the control group (N = 26) who will receive traditional present simple instruction with sage on the stage method. Meanwhile, the experimental group (N = 24) will be trained to use GPT as a learner autonomy enhancement tool, focusing on reducing grammatical errors when using present simple tense.
The researcher would prepare 30 sentences contain grammatical mistakes related to present simple tense in each sentence (see appendix below). The participants would sit in sessions for 9 hours (3 classes for each group). The students in control group would receive traditional methods of teaching present simple (sage on the stage methods). The students in the experimental group would be trained on how to create individual account in GPT portal, how to use facilities of the program, and how to use customization function to give instruction to the software in order to respond to their orders. To help ChatGPT software responds to needed tasks (correcting the grammatical errors) through custom instructions, we give it specific and accurate instructions as follows:
1) Teach EFL learners the rule of present simple tense by correcting the grammatical mistakes in the sentences provided.
2) Feedback should focus only on the rules of the present simple tense.
3) Find the grammatical mistakes related to the present simple tense in the sentences.
4) Underline the grammatical mistakes in the sentences.
5) Give students two chances to correct the grammatical mistakes.
6) Praise their attempt after each attempt.
7) Do not give the correct sentences in the first two chances.
8) If unsuccessful, give the students the correct present simple’ rule in the third attempt.
According to the instructions fed into the ChatGPT application after each attempt, it gives automated feedback and hints for assistance for the next attempt with words of praise. See Figure 1 and Figure 2 below:
Figure 1. A screenshot of the ChatGPT page shows how learners try writing sentences without grammatical mistakes. If they fail, ChatGPT gives them feedback about their mistakes and encourages them to try again. It gives them two attempts.
Figure 2. (A screenshot of the continuous response of GPT to learners); in the third trial if the learner writes the sentence correctly, ChatGPT praises him. If unsuccessful, it gives him the correct sentence with praise to continue the practice.
During the third session, a post-test will be administered to both groups, aimed at assessing the extent to which the LLM-ChatGPT application contributes to the improvement of learner autonomy and the reduction of teacher centeredness. To gather student feedback, a questionnaire will be distributed to the experimental group, seeking insights into how the program might contribute positively to enhance learner autonomy. The questionnaire has 20 statements that are adapted from the Unified Theory of Advance and use of Technology (UTAUT) created by Venkatesh (2003). The massive incorporation of LLM-ChatGPT into the language learning process in general and within this research in particular catalyzes the enhancement of pedagogical methods toward learner autonomy. It helps EFL learners at King Khalid University (KKU) take the initiative towards autonomous learning in all courses.
5. Data Analysis
The data analysis involved utilizing SPSS software to ascertain the means and standard deviations of learners’ achievement in the tests. Additionally, T-tests were employed to scrutinize the scores of the tests within the two distinct groups. Post hoc test was also conducted to know and compare the differences between the groups.
The investigation of learners’ perception of employing GPT as an enforcement approach to enhance learner autonomy was conducted via a questionnaire distributed directly to the participants (experimental group) within the classroom setting. A total of 23 participants responded to the survey items.
6. Result
The results of the T-test (see Table 1) reveal that participants in the experimental group, utilizing LLM-ChatGPT for autonomous learning of the present simple tense rule without teacher intervention, achieved scores with a mean (M) of 13.40 and a standard deviation (SD) of 2.62. Conversely, those who were instructed the present simple tense rule through traditional methodologies with a teacher-centered approach attained scores with a mean of 13.81 and a standard deviation of 2.89. These findings suggest that the LLM-ChatGPT tool fosters learner autonomy and assumes a pedagogical role akin to a teacher’s instruction in teaching the present simple tense. Consequently, the statistical analysis indicates no significant disparity between the two groups. In summary, the comparable scores achieved by both groups in the tests provide evidence that the utilization of LLM-ChatGPT significantly enhances the learner autonomy approach while concurrently diminishing sage on the stage approach.
Table 1. Statistical analysis of T-test for the experimental group.
Spss analysis |
Experimental group = 24 |
Control group = 26 |
M |
SD |
M |
SD |
Pre-test |
9.90 |
3.5 |
8.95 |
2.95 |
Post-test |
13.40 |
2.62 |
13.81 |
2.89 |
7. Questionnaire
The emphasis on learner autonomy has emerged as a focal point in pedagogical approaches. Substantial research has been dedicated to exploring methods to enhance student-centered learning while reducing teacher-centeredness. This study takes a proactive stance by employing LLM-ChatGPT as a means to assess its efficacy in promoting learner autonomy. It substantiates the appropriateness of it, with participants exhibiting positivity about using it. To assess the learners’ attitudes and perceptions regarding the utilization of ChatGPT to enhance the learner autonomy approach, a questionnaire was disseminated to the students comprising the experimental group (See Table 2). It includes 20 statements which are adapted according to the UTAUT-1 model. Its constructs are effort expectancy (1 - 5), performance expectancy (6 - 8), social influence (9 - 10), facilitating condition (11 - 16), and behavioral condition (16 - 20). They are implicitly incorporated in an integrated mode. The cale of Likert is used to refer to the learners’ response; the codes refer to: 1) Strongly Disagree, 2) Disagree, 3) Neutral, 4) Agree, 5) Strongly Agree.
Table 2. Analysis of the questionnaire distributed to the experimental group.
N. |
Statements |
1 |
2 |
3 |
4 |
5 |
M |
SD |
1 |
LLM-ChatGPT application is useful tool to foster learner autonomy |
0 |
3 |
4 |
10 |
2 |
4.18 |
0.755 |
|
16% |
21% |
53% |
11% |
2 |
LLM-ChatGPT applications help me reducing grammatical errors. |
0 |
2 |
2 |
12 |
3 |
4.28 |
0.78 |
|
11% |
11% |
63% |
16% |
3 |
I can use LLM-ChatGPT application to enhancing my other language skills. |
0 |
4 |
1 |
10 |
4 |
4.29 |
0.89 |
|
21% |
5% |
53% |
21% |
4 |
Using LLM-ChatGPT application
increases the chance to score high grades in grammar courses. |
0 |
1 |
2 |
14 |
2 |
4.15 |
0.72 |
|
5.26% |
10.5% |
74% |
11% |
5 |
Using LLM-ChatGPT did not
improve my performance as
autonomous learner. |
11 |
1 |
2 |
4 |
1 |
2.48 |
1.16 |
58% |
5% |
11% |
21% |
5% |
6 |
The use of LLM-ChatGPT
application did not improve my
academic performance. |
6 |
9 |
0 |
4 |
0 |
2.51 |
1.20 |
32% |
47% |
|
21% |
|
7 |
I become skillful in using
LLM-ChatGPT applications. |
0 |
1 |
0 |
15 |
3 |
4.40 |
0.909 |
|
5% |
|
79% |
16% |
8 |
Application of LLM-ChatGPT is
easy to use. |
2 |
0 |
2 |
13 |
2 |
4.46 |
0.86 |
11% |
|
11% |
68% |
11% |
9 |
Teachers and senior students in my department encourage me to use
LLM-ChatGPT application to learn autonomously. |
0 |
0 |
3 |
12 |
4 |
4.0 |
0.80 |
|
|
16% |
63% |
21% |
10 |
LLM-ChatGPT application needs group work working together. |
6 |
8 |
2 |
1 |
1 |
2.60 |
1.11 |
31.79% |
41% |
11% |
5% |
5% |
11 |
I have full control over using
LLM-GPT applications |
0 |
5 |
3 |
6 |
5 |
4.39 |
0.910 |
|
26% |
16% |
32% |
26% |
12 |
I had enough sources to use
this application. |
3 |
4 |
3 |
4 |
5 |
3.66 |
0.90 |
15% |
21% |
51% |
21% |
26% |
13 |
I am aware of how to use
LLM-ChatGPT to maximize
self-dependence learning |
0 |
2 |
2 |
7 |
8 |
4.49 |
0.85 |
|
10% |
10% |
36% |
42% |
14 |
The prerequisite of LLM-ChatGPT
application is to be skillful in
computer skills |
0 |
0 |
0 |
10 |
9 |
4.32 |
0.89 |
|
|
|
53% |
47% |
15 |
I had difficulties in using
LLM-ChatGPT application
because it had complicated tasks. |
13 |
2 |
0 |
1 |
3 |
3.21 |
1.05 |
64% |
10% |
|
5% |
15% |
16 |
LLM-ChatGPT application is
harmonious with other aspects of learning |
1 |
2 |
4 |
6 |
7 |
3.19 |
1.08 |
5% |
10% |
20% |
32% |
37% |
17 |
Using LLM-ChatGPT application
suits my learning style. |
0 |
0 |
0 |
7 |
12 |
3.12 |
1.10 |
|
|
|
37% |
63% |
18 |
It is a good idea to use LLM-ChatGPT to minimize teacher-centeredness. |
0 |
1 |
1 |
1 |
14 |
4.41 |
0.79 |
|
5% |
5% |
5% |
85% |
19 |
Incorporating LLM-ChatGPT
makes learning more interesting |
0 |
0 |
0 |
4 |
15 |
4.41 |
0.912 |
|
|
|
20% |
80% |
20 |
For me, LLM-ChatGPT
application is depressing. |
15 |
4 |
0 |
0 |
0 |
2.50 |
1.14 |
80% |
20% |
|
|
|
Generally, the participants (19) in the questionnaire display positive attitude towards using LLM-ChatGPT to enhance learner autonomy. In statements (1 - 5) which are related to effort expectancy’s construct, the participants indicated with strongly agree that ChatGPT is useful and it foster self-learning of present simple tense in particular and other language skills in general. In performance expectancy construct (statements 6 - 8), they felt that ChatGPT application is easy. They became skillful in its application and ready to use it in other academic activities without teacher’s intervention. In social influence construct (statements 9, 10), they strongly agreed that their senior colleagues and teachers always encourage them to use the application autonomously in all academic activities. Regarding the facilitating condition construct (statements 11 - 16), they confirmed having control over using the application and had the required resources to use it. In the construct of behavioral condition (items 16 - 20), the participant felt that the idea of using ChatGPT as a tool for autonomous learning is stunning and pleasant. The participants’ positive attitude towards utilizing this application establishes a basis for its potential generalization to all language learning processes.
8. Discussion
Regarding the first research question “What is the effect of using LLM-ChatGPT on enhancing learners’ autonomy? Is it more effective than the teacher-centered approach?” The results find that using LLM-ChatGPT has a positive effect on enhancing learner autonomy. Through this application and without teacher’s intervention, EFL learners learn present simple tense rule and structure autonomously and they become able to use it with almost nil grammatical errors. It acts the role of teacher by giving automated feedback, regular assistance, opportunities for regular practice and joyful learning environment. It also reveals that there is a clear positive impact of using LLM_GPT on improving learner autonomy approach. The study aligns with Transform Language model (TLM) by Radford [23] which helps EFL learners to comprehend many language tasks as answering questions, natural language interference and smartly responds to many tasks a human being asks for. It is regarded a good tool that enhances learning language autonomously. It is also consistent with Brown et al. [18] GPT-3 mode in translation and summarization. It improves the learners’ skills in translating and summarizing long texts. It also agrees with Delin et al. (2018) BERT language model which deals with deep language structure. This model is used to achieve tasks concerning natural language understanding and sentiment analysis. There is also alignment of this study with the model of Yang et al. [22] in writing research. This model is used to help learner understands meaning of words in a context and makes him able to consider the relationship between words in a sentence which leads to comprehend the meaning of a whole text. LeCun et al. (2015) used NLP model which allow computer interpreting texts and generating speech. It gives virtual assistance like Siri and Alkex which has the function of understanding and responding to spoken commands. Last but not least, the study agrees with the idea of Livingstone [17] who praises the role of AI models in enhancing learner autonomy. He concludes that the use of high-quality multimedia applications and programs, to a great extent, stimulates and fosters language learning.
Regarding the second research question “What is the learners’ attitude/perception of using LLM-ChatGPT as a tool for autonomous learning?” The result displays a positive attitude toward implementing LLM-ChatGPT in learning present simple tense as a model to fostering learner autonomy. They affirmed that the application is amazing and they benefit from this software to enhance autonomous learning. The result confirmed this attitude and it is aligned with studies mentioned above.
9. Conclusion
The study proves the effectiveness of using LLM-ChatGPT in enhancing learner autonomy and reducing teacher’s centeredness in the learning process. The participants remember the structure of the present simple tense and use it correctly without errors after using this software because it gives them automated feedback, spelling assistance, and opportunities for regular practice. The role of EFL teachers in teaching process is indispensable; however, they may find value in the outcomes of this study and choose to incorporate AI tools into their pedagogical strategies and leverage the findings to bolster learner autonomy proficiencies.
10. Limitations
Throughout the journey of this study, some limitations are observed and are listed below in order to be addressed by scholars who have an interest in this domain. Some of them are:
1) The instructors’ perception is not measured in this study. Therefore; further studies are required to investigate the teachers’ perception of using LLM-GPT chat to maximize learner autonomy.
2) This research focused on present simple tense. In further researches, it is necessary to investigate the effect of using this application to learn all grammatical rules.
Funding
“The author extends his appreciation to the deanship of research and graduate studies at King Khalid University for funding this work through small group research under grant number RGP1/23/45”.
Appendix
The uncorrected sentences which are fed into ChatGPT
1) They works at the same company.
2) I practices yoga every morning to relieve stress.
3) I does not drink coffee every morning.
4) Does they do their homework?
5) She watch Tv in the morning.
6) They played soccer every evening.
7) He read a book before going to bed.
8) The sun do not rises form the west.
9) They takes bus to the work every day.
10) She talk to her friends on the phone regularly.
11) He exercise at the gym three times a week.
12) She do not study for her exams on weekends.
13) He cook dinner for his family every night.
14) She checked her emails first thing in the morning.
15) They goes to the movies once a month.
16) He is reading a book before going to bed.
17) He brush his teeth twice a day.
18) He help his children with their homework.
19) It do not work properly.
20) Does they visit you regularly?
21) We does not eat breakfast in early morning.
22) How often does the old men go to the hospital?
23) The birds sing on the tree every morning.
24) The water boil at 100 degrees.
25) The earth move around the sun.
26) The class start at 8:00 AM every morning.
27) I does not like eating ice cream.
28) We prays 5 times a day.
29) How often do he visit his brother in Jeddah?
30) I help my friend when he need.