A comprehensive Examination of the potential application of

Chat GPT in Higher Education Institutions

1 2
Elham Tajik and Fatemeh Tajik
1
University of Rochester
2
Affiliation not available

October 30, 2023

Abstract
Posted on 12 Apr 2020 | CC-BY 4.0 | https://doi.org/10.36227/techrxiv.22589497.v1 | e-Prints posted on TechRxiv are preliminary reports that are not peer reviewed. They should not b...

The emergence of large language models such as Chat Generative Pretrained Transformer (GPT) has resulted in a paradigm
shift in the education space. GPT, released by OpenAI, can generate suitable responses to human input text in a conversational
style, making it attractive for laborious written tasks, such as summarizing and responding to questions in higher education
institutions. While several studies reveal the real implementations of GPT in higher education institutions, a comprehensive
examination of its potential at the micro and macro levels of higher education is lacking. To address this gap, the researchers of
this study examined the potential capabilities of GPT through three lenses: student-facing, teacher-facing, and system-facing
tools. The purpose of this study is to provide a robust and nuanced understanding of the appropriateness of GPT’s use among
stakeholders, such as educators, administrators, and students in higher education. Finally, the study presents a set of applicable
recommendations for future researchers.

1. Introduction
In late November 2022, the release of the Generative Pretrained Transformer (GPT) caused a sensation.
OpenAI’s CEO, Sam Altman, announced that more than one million users logged into OpenAI to utilize
GPT within a week. ChatGPT builds upon its predecessor, GPT3.5, and is based on a large language model
(LLM) trained on vast amounts of digitized data. The latest version of LLM, GPT, is trained on 175 billion
parameters to generate text that closely resembles human-generated text. GPT is based on a pre-trained
language model that quickly understands human queries and generates seemingly authentic input text (1,
2).
Relatedly, ChatGPT is an artificial intelligence dialogue-based tool capable of generating human-like re-
sponses to human prompts. In addition to working with a variety of languages such as German, French,
Spanish, English, and so forth, it can produce both standard and colloquial language (2). the major applica-
tions of GPT include language translation, text summarization, content generation, code creation, responding
to questions, and composing a story, play, and essay (2-4).
GPT has attracted significant attention from various disciplines, including MBA (5, 6), medical (1, 7),
cognitive science (8), computer science (9), and education (3, 10). In a sense, the emergence of GPT has
brought about significant changes in various fields, including but not limited to education.
While some studies have explored the implementations of GPT in higher education institutions, there is a
lack of comprehensive examination of its potential at the micro and macro levels of higher education (3,
11). This study fills that gap by examining GPT’s through three lenses: student-facing, teacher-facing, and
system-facing tools derived from Baker and Smith terminologies (12). The purpose of this study is to provide
a nuanced understanding of GPT’s appropriateness among stakeholders, such as educators, administrators,
and students in higher education and to offer recommendations for future researchers.

1
 Based on the image presented in Figure 1, we have used DALL-E (13) in order to create an image demonstrat-
ing how GPT could potentially impact stakeholders in higher education. The image portrays the interplay
between GPT and different stakeholders such as educators, administrators, and students.
Posted on 12 Apr 2020 | CC-BY 4.0 | https://doi.org/10.36227/techrxiv.22589497.v1 | e-Prints posted on TechRxiv are preliminary reports that are not peer reviewed. They should not b...

Figure 1. Impact of ChatGPT on students, teachers, and administrators in higher education institutions
(created by DALL-E).
2.1. Student-facing tool
Student-facing tools such as intelligent tutoring systems are being used in education to help students learn
new concepts (12). GPT can be considered a student-facing tool because students can use it to improve
their deep learning, critical thinking and writing skills.
In 1962 Vygotsky proposed that two main concepts, spontaneous and scientific, are responsible for human
cognitive development (14). He emphasized the importance of the relationship between spontaneous con-
cepts, which arise from our lived experiences, and scientific concepts, which are learned from instructional
environments, in the development of higher-order thinking (14). To bridge gap between spontaneous and
scientific concepts, introduced the zone of proximal development, which refers to the difference between what
students can do without assistance (i.e., spontaneous concepts) and what they can do with assistance (i.e.,
scientific concepts) (14). For instance, while students may be exposed to the definition of decimal concepts in
formal instruction, they may struggle to apply their knowledge to real-world situations like calculating a bill
or tip at a restaurant. However, With the help of GPT, students can access unlimited examples of decimals
in various contexts, providing them a deeper and more comprehensive understanding of the concept. Thus,
GPT can aid students in being familiar with the application of new concepts (e.g., decimals) in different

2
 contextualized situations.
Additionally, GPT can also serve as a learning companion. Teachers often ask students to evaluate their
peers’ assignments or tasks which can lead to negative feedback and discord among students. However,
students can evaluate GPT’s output without worrying about offering extremely negative or offensive feedback.
By engaging in back-and-forth conversations with GPT, students can co-create meaning and deepen their
understanding of the concept by bridging the gap between GPT’s correct and incorrect responses. GPT
can provide all individual students with immediate and personalized feedback to help them improve their
understanding of the new concept (10).
The main challenge that students face is the illusion of explanatory , which refers to the belief that students
think that they have a complete understanding of a particular topic while they actually have a shallow
understanding (15). ChatGPT can play an important role in reducing the illusion of explanatory depth by
allowing students to modify GPT’s output step -by -step and create fundamental prompts (e.g., providing
chatGPT a persona) (5). This gives students the opportunity to engage more deeply with the material and
develop a more accurate understanding of the topic.
Overall, while students can benefit from chatGPT, they need guidance on how to work with GPT to fully
Posted on 12 Apr 2020 | CC-BY 4.0 | https://doi.org/10.36227/techrxiv.22589497.v1 | e-Prints posted on TechRxiv are preliminary reports that are not peer reviewed. They should not b...

understand a specific topic by thinking logically and assessing GPT’s output.

2.2. Teacher-facing tool
GPT has numerous potential applications education, including reducing teachers’ workload by automating
tasks such as writing lesson plans, designing class syllabus outlines, creating quiz/test questions, grading
student performance, designing rubrics, and providing direction for learning activities, among others (11).
And also, inexperience teachers can benefit from GPT as a training tool, as it can reduce time pressure
and provide thought-provoking examples and strategies aligned with learning objectives. The GPTTeach
prototype, which enables untrained teachers to practice teaching with “GPT_ simulated students” before
teaching in a real environment, is an example of this (16). In this case, Tate and her colleagues have also
suggested that GPT has the potential to act as a versatile and adaptable teaching assistant that can establish
a natural dialogue with students, creating tutor-tutee relationships in and out of the classroom (3).
The potential application of GPT as a teacher-facing tool can be classified into four key facets of teaching
processes: planning, instruction, assessment, and interaction with other stakeholders such as families and
administrators. GPT can support educators in writing and updating their learning objectives based on
student needs and interests, helping to keep students engaged in a learning environment. With the help of
GPT, teachers can offer timely and constant feedback on students’ writing assignments, facilitating forma-
tive assessment (11, 17). In instruction, GPT can provide personalized learning activities tailored to each
student’s abilities and potential, allowing educators to better address individual students’ needs (11, 17). For
example, GPT can generate learning activities that are tailored to the student’s learning style and interests,
increasing student engagement and motivation. GPT can also provide personalized feedback to students,
identifying areas for improvement and providing suggestions for how to improve their work. Additionally,
AI-generated text can help educators communicate more effectively with parents and administrators, en-
suring that students’ educational needs are met and their learning gains are improved. For example, GPT
can generate personalized messages for parents that inform them about their child’s progress, upcoming
assignments, and any issues the teachers might be facing. This saves time for educators and ensures that
parents are regularly informed and involved in their child’s learning process. Additionally, GPT can generate
reports for administrators summarizing student performance and identifying their strengths and weaknesses,
enabling administrators to proactively act to support students.
Overall, the integration of GPT into the teaching process can significantly reduce teachers’ workload and
allow them to better focus on all individual students’ needs. Figure 2 illustrates this potential using the
example of DALL-E (13) in the teaching process.

3
Posted on 12 Apr 2020 | CC-BY 4.0 | https://doi.org/10.36227/techrxiv.22589497.v1 | e-Prints posted on TechRxiv are preliminary reports that are not peer reviewed. They should not b...

Figure 2. ChatGPT supports teachers in different facets of teaching processes (i.e., assessment, planning,
and instruction) (designed by DALL-E).
2.3. System-facing tool
the application of GPT as a system-facing tool in higher education can streamline administrative tasks,
saving time, and reducing the workload for administrators and managers. One example of the use of GPT
can be to analyze applicants’ written materials, including statements of purpose, letters of recommendation,
and writing samples.by detecting applicants who meet minimum requirements, GPT can assist in screening
and selecting promising candidates, allowing the admissions committee to make more informed decisions and
schedule interviews efficiently.
However, it is essential to recognize the limitations of GPT in system-facing tasks, particularly the potential
for biases, such as stereotypical and social biases (18-20). It is crucial for administrators to be aware of
these biases before implementing GPT in higher education in order to avoid unintended consequences, such
as the unintentional exclusion of marginalized groups from the application screening process. For instance,
when LLMs analyze curricula vitae for recruitment and career guidance, they might be less likely to offer
job opportunities or higher salaries to candidates from ethnic minorities (20). Therefore, it is essential to
mitigate biases and ensure that the use of GPT in higher education is fair, equitable, and inclusive.
3. Discussion and Inclusion
The presence of AI-based technologies represents a significant paradigm shift in the infrastructure of higher
education. When educational systems are open to such drastic change, they would be able to harness
the power of AI-enabled instruments such as ChatGPT in optimizing teaching, learning, and educational

4
 management, which are three indispensable pillars of each thriving institution. For example, at the level
of educational administration, administrators have a lot of time detecting students at risk of dropping out
of courses since GPT can handle a set of tasks such as responding to prospective applicants’ emails or
curating applicant data, remarkably reducing their workload. Similarly, at the teaching level, GPT can help
teachers have more time to assist all individual learners in gaining a robust and systematic understanding
of new concepts by handling tasks like grading and summarizing student works. Furthermore, a study by
Philips et al shows that embedding one of the versions of GPT (GPT3) into an educational game can assist
the instructor in developing their insight into student collaboration in a computer-supported collaborative
learning environment (21). Additionally, at the learning level, many researchers indicate that GPT can have
a remarkable influence on individual learners’ outcomes (3, 6). For instance, GPT can identify a student’s
weakness and create an interactive learning experience in which students can constantly interact with a
virtual peer in a conversational style, facilitating their learning process.
Overall, While GPT has inherent limitations, such as its inability to reason about the physical and so-
cial world, temporal reasoning, factual errors, bias and discrimination, spelling/grammar errors, lack of
self-awareness, ethics/morality, emotions and idioms, transparency, reliability, robustness and security, and
plagiarism (2, 22), it cannot be denied that GPT and other LLM tools, such as T5 (23), BLOOM (24),
Posted on 12 Apr 2020 | CC-BY 4.0 | https://doi.org/10.36227/techrxiv.22589497.v1 | e-Prints posted on TechRxiv are preliminary reports that are not peer reviewed. They should not b...

GPT3 (25), OPT (26), BERT (27), and Roberta (28) are positively revolutionizing the infrastructure of
education at both macro and micro levels. The close interrelationship between humans and LLM (GPT) can
leverage the complementary strengths of humans and AI, leading to a more efficient and effective education
system that benefits all stakeholders involved. Therefore, while it is important to acknowledge and address
the limitations of GPT and other LLMs tools, their potential for positive impact on education cannot be
ignored.
4. Future prospective
Although this study provides a thorough evaluation of ChatGPT, there are some limitations that need to
be addressed in future studies, as outlined below:
Firstly, there is a need for further research to explore the potential of GPT as a system-facing tool for
university staff and administrators, as most of the existing research focuses on student-facing and teaching-
facing tools.
Secondly, a comprehensive examination of the capability of GPT in supporting students with developmental
disabilities, such as a developmental delay in communication and language, is needed.
Lastly, it is unclear to what extent ChatGPT and other LLMs can leverage AI-based technology deployed in
the learning environment, such as educational games and intelligent tutoring systems, to support teachers
and students. Therefore, more research is needed to explore the incorporation of GPT into AI-powered
software and its potential impact on education.
Acknowledgments
We would like to express our gratitude to the anonymous peer-review who provided us with valuable and
helpful feedback during the process of writing this article. Additionally, we utilized pro-chatGPT to assist
us in editing the content of this article.
Conflict of interest
We do not have any conflicts of interest to disclose.
References :
1. Gilson A, Safranek CW, Huang T, Socrates V, Chi L, Taylor RA, et al. How does CHATGPT perform
on the United States Medical Licensing Examination? the implications of large language models for medical
education and knowledge assessment. JMIR Medical Education. 2023;9(1):e45312.
2. Borji A. A categorical archive of ChatGPT failures. arXiv preprint arXiv:230203494. 2023.

5
 3. Tate T, Doroudi S, Ritchie D, Xu Y. Educational Research and AI-Generated Writing: Confronting the
Coming Tsunami. 2023.
4. Williams C. Hype, or the future of learning and teaching? 3 Limits to AI’s ability to write student essays.
2023.
5. Mollick E. My class required AI. Here’s what I’ve learned so far. 2023 [
6. Terwiesch C. Would Chat GPT3 Get a Wharton MBA? A Prediction Based on Its Perfor-
mance in the Operations Management Course. Mack Institute for Innovation Management at the
Wharton School, University of Pennsylvania Retrieved from: https://mackinstitute wharton upenn
edu/wpcontent/uploads/2023/01/Christian-Terwiesch-Chat-GTP-124 pdf [Date accessed: February 6th,
2023]. 2023.
7. Cascella M, Montomoli J, Bellini V, Bignami E. Evaluating the Feasibility of ChatGPT in Healthcare:
An Analysis of Multiple Clinical and Research Scenarios. Journal of Medical Systems. 2023;47(1):1-5.
8. Mahowald K, Ivanova AA, Blank IA, Kanwisher N, Tenenbaum JB, Fedorenko E. Dissociating language
and thought in large language models: a cognitive perspective. arXiv preprint arXiv:230106627. 2023.
Posted on 12 Apr 2020 | CC-BY 4.0 | https://doi.org/10.36227/techrxiv.22589497.v1 | e-Prints posted on TechRxiv are preliminary reports that are not peer reviewed. They should not b...

9. Qin C, Zhang A, Zhang Z, Chen J, Yasunaga M, Yang D. Is chatgpt a general-purpose natural language
processing task solver? arXiv preprint arXiv:230206476. 2023.
10. Pardos ZA, Bhandari S. Learning gain differences between ChatGPT and human tutor generated algebra
hints. arXiv preprint arXiv:230206871. 2023.
11. Trust TW, Jeromie;Mouza,Chrystalla. Editorial: ChatGPT: Challenges, opportunities, and implications
for teacher education. Contemporary Issues in Technology and Teacher Education,. Contemporary Issues in
Technology and Teacher Education. 2023;23(1).
12. Baker T, Smith LS, editors. Educ-AI-tion Rebooted ? Exploring the future of artificial intelligence in
schools and colleges2019.
13. DALL-E: Introducing outpainting Retrieved from
https://openai.com/blog/dall-e-2-outpaint/2022, August,31 [
14. Vygotsky LS. The Development of Scientific Concepts in Childhood. 1962.
15. Rozenblit L, Keil F. The misunderstood limits of folk science: An illusion of explanatory depth. Cognitive
science. 2002;26(5):521-62.
16. Markel JM, Opferman SG, Landay JA, Piech C. GPTeach: Interactive TA Training with GPT Based
Students. 2023.
17. Zhai X. ChatGPT user experience: Implications for education. Available at SSRN 4312418. 2022.
18. Zhuo TY, Huang Y, Chen C, Xing Z. Exploring AI ethics of ChatGPT: A diagnostic analysis. arXiv
preprint arXiv:230112867. 2023.
19. Weidinger L, Uesato J, Rauh M, Griffin C, Huang P-S, Mellor J, et al., editors. Taxonomy of risks posed
by language models. 2022 ACM Conference on Fairness, Accountability, and Transparency; 2022.
20. Weidinger L, Mellor J, Rauh M, Griffin C, Uesato J, Huang P-S, et al. Ethical and social risks of harm
from language models. arXiv preprint arXiv:211204359. 2021.
21. Phillips T, Saleh A, Glazewski KD, Hmelo-Silver CE, Mott B, Lester JC. Exploring the use of GPT-3
as a tool for evaluating text-based collaborative discourse. Companion Proceedings of the 12th. 2022:54.

6
 22. Baidoo-Anu D, Owusu Ansah L. Education in the Era of Generative Artificial Intelligence (AI): Un-
derstanding the Potential Benefits of ChatGPT in Promoting Teaching and Learning. Available at SSRN
4337484. 2023.
23. Raffel C, Shazeer N, Roberts A, Lee K, Narang S, Matena M, et al. Exploring the limits of transfer learn-
ing with a unified text-to-text transformer. The Journal of Machine Learning Research. 2020;21(1):5485-551.
24. Scao TL, Fan A, Akiki C, Pavlick E, Ilić S, Hesslow D, et al. Bloom: A 176b-parameter open-access
multilingual language model. arXiv preprint arXiv:221105100. 2022.
25. Brown T, Mann B, Ryder N, Subbiah M, Kaplan JD, Dhariwal P, et al. Language models are few-shot
learners. Advances in neural information processing systems. 2020;33:1877-901.
26. Zhang S, Roller S, Goyal N, Artetxe M, Chen M, Chen S, et al. Opt: Open pre-trained transformer
language models. arXiv preprint arXiv:220501068. 2022.
27. Devlin J, Chang M-W, Lee K, Toutanova K. Bert: Pre-training of deep bidirectional transformers for
language understanding. arXiv preprint arXiv:181004805. 2018.
28. Liu Y, Ott M, Goyal N, Du J, Joshi M, Chen D, et al. Roberta: A robustly optimized bert pretraining
Posted on 12 Apr 2020 | CC-BY 4.0 | https://doi.org/10.36227/techrxiv.22589497.v1 | e-Prints posted on TechRxiv are preliminary reports that are not peer reviewed. They should not b...

approach. arXiv preprint arXiv:190711692. 2019.