A Comparative Analysis of Grammatical Structure

Between Student-written Vs. AI-Generated Essays

______________________________________________

In Partial Fulfillment of the Requirement

In Technical Writing

________________________________________________

Researchers:

Abringe, Ma. Helen Joy

Fabiilar, Ivy Mae, D.

Odevilas, Joycelyn, C.
 Chapter I

INTRODUCTION

Background of the Study

Lately, the advances in artificial intelligence (AI) had led to several changes

across different sectors especially in education. AI-generated written content, such

as essays is one area that attracted a great deal of interest from the perspective of

both students and teachers.

In the middle of the 21st century, Artificial Intelligence (AI) is a technology that

makes machines perform like humans in this task and developed such devices

greatly. This is done by the fact that it is programmed to simulate the human brain

which makes it accomplish things such as visual perception, speech recognition,

decision making and language translation. The network can accumulate data or

prompts, and then make it visible to the human being. Then, the system is able to

accumulate those data, recognize patterns on them, and make a transformation of its

code in such a way that is never been tried in the traditional way of programming a

computer (International Business Machines (IBM), 2020).

Artificial Intelligence (AI) is born in 1956 at a Dartmouth College, Hanover,

New Hampshire conference. John McCarthy from MIT was the father of Artificial

Intelligent, Marvin Minsky from Harvard, IBM's Nathaniel Rochester and Claude

Shannon of Bell Labs in New Jersey provided the intellectual stimulus for this get-

together were the other three. Meanwhile, the brilliant inventor Alan Turing is

commoditized by the concept AI who got the inspiration from his creation of a

“universal machine” in the late 1940s a digital computer that can be programmed to
simulate any other digital computer or other discrete-state machines that can do

more number-theoretic calculations (Turing 1950; McCorduck 2004).

According to an article published in Techtarget.com, titled 'Features of Artificial

Intelligence' Artificial Intelligence have the ability to learn and adapt, understand and

generate human language that are based on natural language processing or the

NLP, interpret information from visual images such as computer vision. AI can also

perform automation and machine learning. This means that AI can function without

humans’ help, understand and generate human language, read visual information,

and find easier solutions for challenging tasks. However, one of its common

applications is using it as a writing assistant, where intelligent software helps with

content generation, grammar and writing style-checking, and improving the quality of

written text especially in writing essays (Zhang et al., 2020).

The use of AI writing tools is a trend that is not only popular in particular

localities but is widespread in the world. After a survey had been done by Gartner

(2022), it was established that 25% and about 30% of teachers and students

respectively worldwide are using AI tools in writing and other applications to meet the

educational needs. The increasing use of AI tools in the education sector is observed

not only in high education but also in ESL writing and school administrative works

through AI tools and technology. The implementation of AI tools in education is also

seen to be rising in different countries including the Philippines. One other

experiment was also executed by the Department of Education (DepEd, 2023)

where-in results presented revealed that nearly fifteen percent of teachers and

twenty percent of students in urban areas are currently employing AI tools for

learning purposes, particularly in writing activities. Furthermore, the scale of users is

forecasted to rise due to the fact that it is not only available but also easily

accessible.

AI writing tools are commonly used by teachers and students results in many

uses that they might have them for. AI tools enable users to save time by since they

can easily generate content in the writing process and give quick feedback that will

result in efficiency and time management which the others (Smith & Anderson, 2021)

report. These tools also make real-time grammar and style corrections, which in turn

ensure better-quality writing (Jones, 2021). Students personally benefit from AI tools

as they are used as tutors, with individuals getting feedback and suggestions to

improve their writing from them (Brown, 2022). Also, AI writing tools usually are more

accessible and affordable than traditional educational assisting tools (Lee & Kim,

2022). AI tools are the main resources for teachers to evaluate the written work of

their students in a consistent and accurate way. They also become the best way to

improve students' grammar accuracy and writing skills among as it enables them to

write freely from grammar inaccuracies and errors with the assistance of existing AI

grammar checker tools such as Grammarly and Quill Bot. There were also AI writing

tools that can help generate written text and essays such as Chatgpt, Cici, Gemini,

etc.

However, this rise also to the question: are AI-generated essays more or less

accurate grammatically than human-written student written essays? The use of

Artificial Intelligence, particularly writing tools in educational settings is a well-

discussed topic that covers large researches and studies that examines both of its

strengths and limitations (Smith, 2020; Johnson, 2021). AI generated content,

especially essays, are highly advantageous and offers not only a level of consistency

but also the "already increased in production" and has the ability to write text without
error. But this also rise to another question, 'How credible are AI generated essays?

Are they really free from grammatical errors?'

This study aims to answer these questions by comparing the results on the

frequency of grammatical errors present in both AI-generated essays and student-

written essays. By examining the grammatical structure of student-written essays

and AI-generated essays, the study will provide students with an evaluation of their

writing and also feedback towards AI writing tools to know both of their grammatical

accuracy (Miller, 2018; Thompson, 2019). This research also will add to the literature

of previous researches in integration of AI writing tools in education and for students'

responsible utilization of these tools in writing their essays.

Statement of the Problem

The study attempts to analyze essay content of student-written and ai-

generated essays. It seeks to determine the answer to the research objectives:

1. What is the frequency of grammatical errors present in High School Students'

written essays and AI (GPT- 3.5) essays in terms of Grammar Error Types?

Subject-Verb Agreement

Preposition

Verb Tenses

Conjunctions

2. How does the frequency of grammatical errors present in essays High School

Students' written essays compare to the essays written by AI in terms of Grammar

Error Types?

Subject-Verb Agreement

Preposition

Verb Tenses

Conjunctions
Significance of the Study

Students

The results of this study can help students develop better skills especially in

writing essays. By being aware of the common grammar errors they commit in their

essays, they can use it in improving their vocabulary skills.

Educators

By measuring the frequency of errors in student versus AI-generated essays,

educators will be able to develop a better understanding for where current AI tools

might excel or fall short when it comes teaching writing skills.

Educational institutions

Universities and Schools can utilize the findings to decide on how AI tools

should be used in education, but most importantly making sure that these are

applied effectively to further improve student learning results.

AI Developers

AI developers can also benefit from the results of this study. It can be used to

increase accuracy and dependability of services offer to create more credible

generated content.
Scope and Delimitation

This study focuses on a comparative analysis of grammatical errors in

essays written by high school students and those generated by an AI model (GPT-

3.5). The participants of the study will include 20 high school students, they will be

selected based on their availability. The participants will be given one hour to write

their essays based on the given writing prompts. The 20 AI-generated essays will be

generated in GPT-3.5 with the same essay prompts assigned to the students. The

study will examine the frequency of grammatical errors occurred in both sets of

essays. The researcher will ensure reflexivity and ethical consideration throughout

the data gathering process.

Definition of Terms

1. Student written essays

Operational Definition: Student Written Essays are essays that is written by students

without the help of any AI writing assistance tools. This is the output that will be

collected from the 20 high school students’ participants.

Conceptual Definition: Student-written that are typically given by teachers to know

their level of learning in terms of non-oral communication such as wtiting essays.

(https://www.verywellmind.com/how-to-write-an-essay-in-apa-format-2794837)

2. AI-Generated Essays

Operational Definition: AI Generated Essays are essays written by any AI writing tool

assistance such as the most use ChatGPT models.

Conceptual Definition: These essays are generated using modern technology and

natural language processing techniques to generate text and on a given sample

prompt. This is the output that will be collected for AI generated essays from the

model GPT. (https://www.forbes.com/sites/forbestechcouncil/2021/09/09/how-ai-is-

transforming-the-writing-industry/?sh=6a0b9a3d6f8c)

3. Grammatical Errors

Operational Definition: These are grammar errors committed in both essays

including issues in usage of some parts of speech.

Conceptual Definition: These errors can include grammatical mistakes in subject-

verb in some parts of speech (https://www.grammarly.com/blog/grammar-checker/)

 Chapter II

REVIEW OF RELATED LITERATURE

Artificial intelligence or AI in education has both as advantages and

disadvantages especially those students who use it in writing their essays. The

associated literature review goes on to make a comparative analysis between the

grammatical errors in essays written by students and those by AI, looking at the

frequency of grammatical errors that have occurred at both essays. The following

literatures include the topics about, AI in Academic writing, Adoption of AI tools by

High school students, and limitations that are present in AI wtiting tools itself that

could possibly the reason why it can also commit some grammatical mistake just like

humans.

Grammatical Structures

Following grammar is the basis of a successful writing. The team of

Grammarly tells us that grammar is a set of rules and standards that govern words,

parts of speech and clauses constructs in any language. The principle of universal

grammar is the sequence of different language elements in sentences, such as the

subject-version, objects-adverbial and others. That gives a sentence its meaning.

SentenceStructure.org (2023) has an opinion that grammatical structure, a way that

a sentence positions all nouns, verbs, adjectives, and adverbs depending on the

structure of sentence. Another, it helps in English Grammar, the term is part of

speech and also includes such things as proper syntax, morphology, and word
choice of grammatical items to convey the message in a clear and understandable

way.

Subject-verb agreement is the most significant language usage rule in

English, which is necessary to a grammatically correct sentence. Johnson (2021)

according to him subject-verb awkwardness can be caused by the lack of

understanding of the whole sentence. For example, "The cat runs" and "The cats

run" are different ways to write the critters' fun (Johnson 2021). Smith (2021) stated

that the implementation of correct subject-verb agreement is vital with regard to the

readability and absorption of knowledge, thus the texts would become more

acceptable to readers.

Furthermore, the choice of a preposition is considered as a decisive factor in

the construction of grammatically correct sentences. The role of prepositions is to

show the relationship between different parts of a sentence and if they are not used

correctly it may lead to the confusion (Smith, 2019). In the same way, Johnson

(2019) mentions that unless we are good at this aspect, the wrong use of

prepositions can bring about ambiguity and/or incorrect readings while the sentence

is being read.

The verb tenses which are used to refer to an action or a subject being a

certain condition ate the other basic element of this. Saying that "Verb tenses have

to be consistent to avoid misunderstandings which may arise if a given paragraph

comprises a sentence with different tenses in it" is what Brown (2020) is focusing on.

With the right use of verb tenses, the situation makes understanding and positioning

of events without
 Conjunctions are also those words that writers use to connect sentences and

clauses together. Thompson (2020) explains the necessity of tracking students’

incorrect use of conjunctions in this way since it will enable them not to commit these

errors anymore. Adams (2020) rebuts the fact that using conjunctions wrongly can

lead to the total idea becoming irrationally makes the narrative unclear to the

readers.

To adhere with the general grammatical rules would include, the pre-

established subjects, verbs, prepositions, verb tenses, conjunctions, etc. Green

(2022) mentions that the fluency of the given rules by the writer improve the

readability and the impression of professionalism of written communication. He also

advises that these grammar rules, when they are commonly known, will result in the

more scientific and organized way of communication.

AI in Academic Writing

AI tools, such as the most common one, OpenAI's GPT-3.5, have gained

popularity for their capacity to produce text that closely copies human writing.

According to Brown and Garcia (2020), these models are developed to do heavy

tasks, giving them the privilege to write essays on a wide range of topics with a high

degree of coherence and grammatical accuracy. However, despite of its evident

benefits and uses, there are times that these texts can also contain grammatical

errors just like how human can also commit grammar mistakes. According to Chen

and Williams (2022), shows that while AI can produce high-quality content, it can still

be 100 percent accurate since some of these generated texts might have an

opposite context that is far not related to the prompts.

 The utilization of Artificial Intelligence (AI) in academic writing, especially in

essay composition, has advanced and become one of the most discussed subjects

recently. AI tools could really take academic writing to the next level. In that regard,

Lewis (2020) conducted a survey that stated that students employing AI-assisted

writing tools were able to finish their essays 30% quicker than those who didn't which

evidenced the time-saving benefits of using these tools (Lewis, 2020).

AI tools can even do things such as providing feedback that is personalized to

a writer's style and needs. Johnson (2022) comments that these tools can change in

response to the way the author writes, thus giving reliable advice on how to improve

your skill (Johnson, 2022).

In addition to the benefits, ethics is another issue to be concerned with.

Williams (2023) raised the issue of AI oversubscribing in their speech during an

educational conference, mentioning that it would defeat the purpose of students'

writing abilities and academic integrity if responsibility is not properly enforced

(Williams, 2023).

Reasons for AI Adoption by High School Students

High school students have increasingly been resorting to the use of AI writing

tools and there is no shortage of reasons behind this. One such reason is that these

tools give instant feedback, which is a must-have for learning and development. With

prompt feedback, students can quickly identify their mistakes and correct them

hence increasing efficiency in learning and results (Smith, 2020; Thompson, 2019).

According to EdTech Magazine survey findings, 73% of the students claim that AI
tools enhanced their understanding of subject matter by providing timely feedback

(Johnson, 2021).

Moreover, AI generated content tutors’ learners on grammar as well as the

structure through which they get to learn about proper grammar and how to construct

correct sentences. Grammarly and ProWritingAid are among platforms with AI

writing tools that offer advice on correcting grammatical issues, improving style or

enriching vocabulary thereby serving as useful examples of applied learning. A study

published in the Journal of Educational Technology found that during a semester

period 67% of students using AI writing tools reported an improvement in their writing

skills (Lee, 2022). These instruments also facilitate learning-by-doing since students

can see their errors immediately and understand what corrections are being made.

Moreover, AI tools also help in saving time and reducing the cognitive load

associated with writing hence allowing learners to concentrate on other academic

work or extracurricular activities (Brown, 2019; Garcia, 2020). AI tools automate

routine tasks like grammar checking and formatting thus freeing up time for students

to focus on more complex and creative aspects of their assignments. Based on a

survey by The Chronicle of higher education, 58% of students acknowledged that

these technological tools facilitated effective management of their academic

workload (Davis, 2021). This is specifically important for high school students who

have several duties to attend to frequently.

Additionally, AI tools inspire customized learning since they adjust easily

according to the unique needs and learning approaches employed by each student.

This approach maps well with Constructivist Learning Theory which advocates for an

individualistic form of education (Jones, 2019). For instance, a study done by

Computers & Education revealed that seventy-two percent of students using AI-

driven personalized learning platforms affirmed they had a clear understanding of

course concepts (Wilson, 2021). Adjusting themselves to student’s progress and

providing personalized feedback makes these platforms highly effective in teaching.

Furthermore, collaborative learning by students is facilitated through AI tools

which step up communication and coordination among them. In that regard, it can be

used to match pupils for assigned group work projects determined by their abilities

and the way they learn. This fosters greater cooperation between peers (Clark,

2020). According to a survey conducted by the International Journal of Educational

Technology in Higher Education, 65% of students said that AI had helped them

collaborate better on group projects (Nguyen, 2022). The ability to foster team

working is invaluable in today’s education context where group activities are

emphasized.

In addition to that, AI makes learning more accessible for students with

disabilities. For example, tools such as speech-to-text or text-to-speech help visually

or physically disabled students access study materials easily (Adams, 2020). In line

with this report released by National Center for Learning Disabilities, around seventy

percent of individuals who have challenges related to reading found themselves

benefiting from using AI tools (Harris, 2021). These technologies guarantee that all

learners including those with physical or instructional troubles can succeed

academically.

Grammatical Errors in Essays Written by Students

 Most student writings, especially in educational settings, are found to be full of

grammatical errors. In this category fall such errors as incorrect verb tenses, tenses

which disagree with the subject, punctuation errors, misapplication of articles, and

prepositions (Thompson, 2019; Williams, 2022). The explanations that have been

advanced for them making these mistakes include low language proficiency, lack of

practice, and poor feedback (Miller, 2018; Brown, 2019.)

On top of the list among the causative factors is limited language proficiency,

especially in ESL students. Anderson, in his research, claims that 65% of ESL

students can be found making grammatical errors frequently due to a lack of

command over the language. In this regard, Lee displaces this with the findings that

70% of grammatical errors existed in student essays because of a lack of English

grammar rules.

Another important factor is poor feedback. The students should be in a

position to realize their mistakes and know how to correct them in time. Dr. Harris

says that 60% of the participants who responded to his study reported that feedback

they got on essays was too general and hence not detailed enough to help them

improve their grammar. Equally important, Garcia (2020) spent much time on

detailed feedback, commenting that students who received comprehensive feedback

improved by 30% in grammatical accuracy in one semester.

Such error patterns are useful in the design of targeted interventions for

bettering student writing. For example, Chen (2020) argues that grammatical errors

can be significantly reduced by eliciting tailored grammar instruction based on

student needs. More than this, AI tools could also contribute by providing immediate

feedback and additional practice. According to a study by Johnson, 2019, students

who used AI-based grammar tools reduced their grammatical errors by 25% in

comparison with peers who were reluctant to use them.

Moreover, peer review and collaborative writing activities can enhance

grammatical competence. According to Clark, students participating in peer review

sessions lessen grammatical mistakes by 20% in the final essays. The activity

assists students in learning from one another and becoming better at critical thinking

and self-evaluation. More than that, it is possible to embed some exercises on

grammar within the writing curriculum. Adams 2021 discovered that those students

who practiced grammar exercises as part of their writing curriculum had 18% fewer

grammatical mistakes. The exercises strengthen the rules of grammar and give the

students the practice needed to apply those grammar rules properly.

Grammatical Errors in AI-Generated Essays

Despite of many services offered by AI, it is still easily fall prey to such

grammatical errors. Mistakes are found in AI-generated text processing due to

complexities involved in NLP and limitations of train data. Mistakes in AI-generated

essays include misused words, awkward phrasing, and syntactic anomalies.

One of the central factors causing these errors is the complexity of natural

language. Even with their power, AI models can sometimes fail to pick up very fine

distinctions and context-dependent meanings convincingly realizable by a human.

According to Xu et al. (2022), it was found that 15% of the sentences generated by

AI models were awkwardly phrased, hence decreasing the readability of the text.

Similarly, Liu and Qian (2021) reported that 12% of AI-generated texts exhibited

inappropriate word choice dues to the inability of the model to understand completely
the contextual intricacies. The restrictions in training data also play a part in

grammatical errors in AI-generated texts. AI models are trained on extensive

datasets, but these datasets may not encompass all linguistic variations or may

include erroneous examples. For instance, a study by Zhang and Wang (2022)

revealed that AI models trained on datasets with inconsistent grammatical structures

resulted in a higher occurrence of grammatical errors. This emphasizes the

significance of high-quality, meticulously curated training data.

Another challenge for AI is the insufficiency of real-time contextual

understanding, whereby AI lacks the human capacity for adaptation of use of

language with respect to the immediate context of discourse or writing. According to

Chen et al. (2021), this mostly causes syntactic anomalies and awkward sentence

structures.

These errors show the crucial, continuing process of developing AI

algorithms and the fact that human supervision plays a paramount role in academic

work. Human proofreading will be one way of verifying that the text generated by

algorithms is correct and coherent. In a survey conducted by EdTech Review, 68%

of educators replied that the content generated by AI should be supervised by

human eyes for the sake of assurance regarding its reliability and correctness.
 REVIEW OF RELATED STUDIES

According to the study titled, 'AI is changing scientists: Understanding of

language learning – and raising questions about an innate Grammar' by Morten H.;

Pablo. C (2022) shows results that even if Artificial Intelligence choice of words is

sometimes strange, nonsensical or contains racist, sexist and other harmful biases,

one thing is very clear that the overwhelming majority of the output of these AI

language models is grammatically correct. They added that, even though there are

no grammar templates or rules hardwired into them, their system relies on linguistic

experience alone, messy as it may be. One of their examples is using the GPT-3.

Wherein they tried to generate a text using a prompt, according to the results, the

GPT-3 can generate believable text reacting to the sample prompts such as “A

summary of the last ‘Fast and Furious’ movie is…” or “Write a poem in the style of

Emily Dickinson.” Moreover, GPT-3 can respond to SAT level analogies, reading

comprehension questions and even solve simple arithmetic problems, all from

learning how to predict the next word.

 This was supported when similar research published in Nature Neuroscience

demonstrated that these artificial intelligence with the features deep-learning

networks seem to use the same computational principles as the human brain. The

research group, led by neuroscientist Uri Hasson, first compared how well GPT-2 – a

“little brother” of GPT-3 – and humans could predict the next word in a story taken

from the podcast “This American Life”: people and the AI predicted the exact same

word nearly 50% of the time.

One more research, which has been carried out in the field of the comparison

of AI writing tools with people in writing quality, was also put forward. The research is

called "Differences between AI and Student-written Essays" received by Rereply.

(2024) .Nadiia Mykhalevych (Mykhalevych) found that these tools exceed in

accuracy by evaluating over 12,000 essays and producing an analysis of the most

repetitive words, phrases and errors. The findings are composed of the writings

made by students (awarded from the writing database IvyPanda) and those

computer-generated by ChatGPT.

According to the results, Student-written essays were far more likely to feature

mistakes, with 78% of essays containing at least one error, compared to only 13% of

AI essays. The most common punctuation errors in student essays included missing

commas, hyphen errors, and missing spaces.

The most common words for students were ‘people’, ‘also’, ‘one’, and

‘however’ while ‘social, ‘cultural’, ‘individuals’, and ‘health’ were used the most

frequently in AI essays. ChatGPT uses 35% less unique vocabulary than humans

with only 7,308 unique words overall, compared to 11,248 unique words from
student essays. Humans are more likely to use common nouns and verbs like ‘one’

or ‘people’, while ChatGPT favors complex and specialized terms like ‘cultural’ or

‘economic.

The results of the study found a huge difference in the overall number of

errors when comparing human writing with AI. With over 45,000 errors across 12,000

student-written essays, there were only 15 overall errors in the 115 ChatGPT-

produced essays. That means while there’s an average of three errors per essay in

student-written work, there was only one error in every ten essays written by AI.

However, this was debunked after a study conducted by the University of

Barcelona showed different results. A research team from the URV which consist of

Humboldt-Universität de Berlin, the universitas Autònoma de Barcelona (UAB) and

the Catalan Institute of Research and Advanced Studies (ICREA), took part in the

project. To compare the accuracy and fluency of the language systems, the

researchers had a competition. The task was for the human subjects and the three

best large language models to write the same text, which was then grammatically

filtered out by the necessary devices. Two of them are GPT3 based and one

(ChatGPT) is under GP3.5. They were posed with a large number of sentences all of

which the subject/verb agreement was the main issue. Among the various sentence

types, there were a simple question, "Is this sentence grammatically correct?” which

was put to both human experiments and language models.

The results showed that humans answered correctly, while the large language

models gave many wrong answers. In fact, they were found to adopt a default

strategy of answering "yes" most of the time, regardless of whether the answer was

correct or not.
 The result is surprising, since these systems are trained on the basis of what

is grammatically correct or not in a language," explains Vittoria Dentella, researcher

from the Department of English and German Studies, who led the study. Human

evaluators train these large language models explicitly about the grammaticality of

the constructions they may encounter.

One researcher named Evelina Leivada, ICREA Research Professor from the

UAB's Department of Catalan Studies added, " Developing useful and safe artificial

intelligence tools can be very helpful, but we need to be aware of their shortcomings.

Since most AI applications depend on understanding commands given in natural

language, determining their limited understanding of grammar, as we have done in

this study, is of vital importance."

"These results suggest that we need to critically reflect on whether AIs really

have language skills similar to those of people," concludes Dentella, who considers

that adopting these language models as theories of human language is not justified

in their current stage of development.

With all of these existing studies, there’s still an insufficient result whether AI-

Generated Essays are more grammatical accurate compares to Student-written

Essays or vice versa. Thus, this study seeks to provide more results and

understanding towards the same research objectives.

 THEORETICAL FRAMEWORK

Error Analysis in Second Language Acquisition (SLA)

According to Corder (1974) and Dulay & Burt (1974), the use of error analysis

frameworks can be a great tool to figure out how learners (in this case, high school

students) take on the learning of language and also the errors that they are

commonly exposed to. The said framework gives its role in the classification and,

further, gaining an in-depth perspective at common grammatical errors, for example,

the misuse of the subject-verb agreement, proper use of prepositions, verb tenses,

and conjunctions in student-written essays.

Example: Corder (1974) remarked error analysis that gives a clue of the

development of grammar due to the systematical errors made by the learners.

Through this knowledge of learners, the acquisition process using specific elements

in the grammar is being shown as well.

Natural Language Processing (NLP) and Error Detection in AI

Some AI-nlp frameworks, in specific, those concentrating on error detection

and correction in AI-generated text, offer a decision to check grammatical errors

made by different types of AI entities such as GPT-3.5. Ways are the line of rule-

based approaches and the design of machine learning algorithms tailored to

detecting errors in syntax and semantics.

Example: Studies on NLP, the most notable of which have been the ones by people

like Sakaguchi et al. (2016), reveal that there exist viable ways of error detection in

machine-generated texts. Also, these methods can be used to analyze the

grammatical errors that AI essays have.

How these Theoretical Frameworks can be used in the study?

The researcher in the process of categorizing these errors should employ SLA

frameworks to do so, that is, they will undertake the classification and quantification

of errors in student-written essays (e.g. subject-verb agreement, prepositions) and

subsequently compare the mistakes that are found in the AI-produced essays.

The researcher will also use NLP framework to quantify and compare the

frequency of grammatical errors and then will compare the student essays to the

ones written by AI in the different types of errors, such as subject-verb agreement,

prepositions, verb tenses, and conjunctions.

 CONCEPTUAL FRAMEWORK

Student-
Written Essays

GRAMMATICAL
STRUCTURE
AI (GPT-3.5)-
Generated
Essays Subject Verb-
Agreement

Prepositions

Verb Tenses

Conjunctions
 HYPOTHESES

Null Hypothesis (H0)

There’s no significant difference between High School Students’ Written

Essays and AI (GPT-3.5) – Generated Essays in terms of the frequency of

grammatical errors (subject verb-agreement, prepositions, verb tenses,

conjunctions).

Alternative Hypothesis (H1)

There’s a significant difference between High School Students’ Written

Essays and AI (GPT-3.5) – Generated Essays in terms of the frequency of

grammatical errors (subject verb-agreement, prepositions, verb tenses,

conjunctions).
References:

Adams, R. (2020). The role of conjunctions in ensuring coherence in writing. Journal

of English Grammar and Usage, 15(2), 45-56.

Brown, M. (2022). AI in Education: The Impact of Artificial Intelligence on Learning

and Teaching. Education Journal, 45(3), 234-245.

Brown, T. (2020). The Importance of Consistent Verb Tenses in Writing. Grammar

Today Journal.

Corder, S. P. (1974). Error analysis and interlanguage. Oxford University Press.

DepEd. (2023). Survey on the Use of AI Tools in Education. Department of

Education, Philippines.

Dulay, H., & Burt, M. K. (1974). Errors and strategies in child second language

acquisition. TESOL Quarterly, 8(2), 129-136.

Gartner. (2022). AI Adoption in Higher Education: A Global Survey. Gartner

Research.

Green, L. (2022). Enhancing Readability through Grammatical Accuracy. Linguistics

Review.

Johnson, M. (2021). Understanding Subject-Verb Agreement. Language Insights.

Johnson, P. (2022). Personalized Feedback through AI in Writing Tools. Educational

Technology Review.

Lee, S., & Kim, J. (2022). Accessibility and AI: Bridging the Gap in Education.

International Journal of Educational Development, 30(4), 315-329

Sakaguchi, K., Post, M., & Berg-Kirkpatrick, T. (2016). Efficient unsupervised

grammar induction for morphologically rich languages. In Proceedings of the 54th

Annual Meeting of the Association for Computational Linguistics (Volume 1: Long

Papers) (pp. 1296-1305).

Smith, A. (2019). The Role of Prepositions in English Grammar. Linguistic Studies.

Smith, A., & Anderson, J. (2021). Efficiency in Education: How AI Tools Save Time.

Educational Review, 48(1), 85-102.

Thompson, D. (2020). Effective Use of Conjunctions in Writing. Writing Techniques

Journal.

Turing, A. (1950). Computing Machinery and Intelligence. Mind, 59(236), 433-460.

Vittoria Dentella et al, Systematic testing of three Language Models reveals low

language accuracy, absence of response stability, and a yes-response bias,

Proceedings of the National Academy of Sciences (2023). DOI:

10.1073/pnas.2309583120

Links:

https://www.sentencestructure.org/how-to-check-grammatical-structures-in-english/

#:~:text=Grammatical%20structure%20in%20English%20language%20is%20simply

%20the%20arrangement%20of,of%20grammar%20and%20sentence%20structure.

https://www.google.com/url?sa=t&source=web&rct=j&opi=89978449&url=https://

www.grammarly.com/blog/sentence-structure/%23:~:text%3DSentence

%2520structure%2520is%2520the%2520order,the%2520use%2520of%2520proper

%2520grammar.&ved=2ahUKEwjXxJy8wLCHAxUfr1YBHe0nCicQudELegQIHhAC&

usg=AOvVaw3_eSfnlXJhKcIMvxe9H9wW

https://preply.com/en/blog/ethical-essay-etchings/?

fbclid=IwZXh0bgNhZW0CMTEAAR0nimYf7xYPTV1dGXy3kCWoO5XzzSXzsTKOD5

1nuVvHajNaKctz_6yK7UY_aem_Sz4UiGP78Hzhj4BxJmtSQQ

https://techxplore.com/news/2024-01-artificial-intelligence-grammar.html#:~:text=A

%20study%20of%20UAB%20and%20URV%20researchers%20and,the%20three

%20best%20large%20language%20models%20currently%20available.

https://theconversation.com/ai-is-changing-scientists-understanding-of-language-

learning-and-raising-questions-about-an-innate-grammar-190594