Available via license: CC BY 4.0
Content may be subject to copyright.
Vol.:(0123456789)
Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w
Discover Education
Review
Exploring theopportunities andchallenges ofChatGPT inacademia
IyolitaIslam1· MuhammadNazrulIslam2
Received: 29 September 2023 / Accepted: 19 March 2024
© The Author(s) 2024 OPEN
Abstract
The Articial Intelligence (AI) natural language model ChatGPT (Chat Generative Pre-trained Transformer), often referred
to as ChatGPT-4, has a wide range of possible uses in the elds of research, business, academia, health, and similar elds.
This language model can accomplish a number of academic jobs that were previously completed by people, taking a
signicant amount of time and eort. The purpose of the article is to investigate ChatGPT’s potential opportunities and
challenges in academia. To attain this objective, a review of relevant literature and online resources (news, articles, etc.)
was carried out. The noticing-collecting-thinking approach was adopted to explore and categorize all observed concerns.
The outcome of this study reveals that research, education, personal skill development, and social aspects constitute
the four broad perspectives that articulate the opportunities and constraints of ChatGPT in academia. For instance, from
theeducation perspective, ChatGPT can help students have a personalized learning experience. On the other hand, it
might provide false information as well as lack the ability to generate responses on its own because those responses
depend on training datasets, which may contain errors. Similarly, from the point of view ofthe personal skill development,
this model may impair a person’s capacity for critical thought and content production; while providing reading and writ-
ing practice sessions and relevant content, it can improve a person’s language prociency.
Keyword ChatGPT, GPT-3, GPT-4 academia, Challenges, Opportunities, Conceptual analysis, Articial Intelligence,
Education, Writing
1 Introduction
ChatGPT is an articially intelligent chatbot initially based on GPT-3, a Natural Language Processing (NLP) model based on
deep learning algorithms [1, 2]. GPT-3 is a multi-modal machine learning model trained on a large text data set to gener-
ate human-like text [3]. This Large Language Model (LLM) can perform several tasks like language modeling, language
translation, and generating text for applications or chatbots (i.e.: ChatGPT) with more reliability and creativity. ChatGPT
was developed by a San Francisco-based Articial Intelligence (AI) research and deployment organization OpenAI [4,
5]. Currently, ChatGPT is freely accessible through the OpenAI web portal. Recently, OpenAI has launched a new version
GPT-4 which can work with images also [6]. GPT-4 is more powerful with lots of new features and fewer mistakes than
the previous version. This model can create, modify, and collaborate with users on creative and technical writing assign-
ments, such as music composition, scripting lms, and adapting to a user’s writing tone [7]. The objective of developing
ChatGPT was to assist users as a dialogic agent in natural language providing helpful and accurate information like
* Muhammad Nazrul Islam, nazrul@cse.mist.ac.bd | 1Department ofComputer Science andEngineering, Bangladesh University
ofProfessionals (BUP), Dhaka1216, Bangladesh. 2Department ofComputer Science andEngineering, Military Institute ofScience
andTechnology (MIST), Dhaka1216, Bangladesh.
Vol:.(1234567890)
Review Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w
customer service, chatbot, and alike [1]. For this, ChatGPT was trained with a massive dataset sourced from the internet.
This dataset includes a wide range of text genres, such as news, articles, scientic papers, social media posts, etc. in dif-
ferent topics and languages [8]. The exact size and composition of the training dataset can be estimated to be on the
order of hundreds of gigabytes or more [9]. The goal of the training process is to learn patterns and relationships within
the text data that allow the model to generate human-like responses to various questions and prompts [10].
Several existing LLMs include BERT [11], Transformer-XL [12], Megatron [13], Jurassic-1 [14], ELECTRA (Eciently Learn-
ing an Encoder that Classies Token Replacements Accurately) [15], Gopher [16], etc. These LLMs were developed to
understand and generate natural language [17]. All of these existing language models can analyze task-specic data,
generate a response, and ne-tune the generated output. These models can be instructed to generate outputs according
to a user inquiry. These previous LLMs were only able to handle one-time probing and also generate responses with a
large variation in case of similar instructions [18]. Usually, LLMs like GPT-3 and GPT-4 can work with more than a hundred
billion parameters.
Being a smart language model with an IQ score of 147 [19], ChatGPT can perform a wide range of functions in the
academic sector, for example, generate, alter, and optimize creative and technical writings based on the context and
mode (formal, informal, etc.) by enhancing the quality, interactivity, and accessibility to education. On the other hand,
ChatGPT can create some challenges in the academic sector also including generating biased output, diculties in ensur-
ing academic integrity, reducing individual writing skills, having limited abilities being a machine, etc. Additionally, there
may be concerns about reducing personalization and the potential for job loss in academia. As such, the motivation of
this research is to address the implications (positive and negative) of ChatGPT in the academic domain.
Since academia is continuously progressive, adapting innovative technologies may enhance the teaching and learn-
ing experience signicantly. By investigating the challenges and opportunities of ChatGPT in academia, this research
aims to contribute to improving the quality of education with eective and ecient use of this AI model. Therefore, the
objective of this research is to explore the insights of ChatGPT in academia in terms of opportunities and challenges
assisting students, teachers, and researchers. To attain this objective, data collected from dierent sources were analyzed
following the noticing-collecting-thinking model. Then, the impacts of ChatGPT in academia were derived as opportuni-
ties and challenges.
The signicance of this study is that it aims to clarify the impacts of using ChatGPT and similar AI-based natural lan-
guage models in learning environments. The purpose of the research is to signicantly contribute to the state of knowl-
edge on AI in education by investigating possible benets, and concerns of this technology in educational contexts.
Again, this study intends to ll a theoretical gap in the literature by focusing on the practical implications of ChatGPT in
educational environments. Although a lot of research has been conducted regarding the general applications of AI in
education, there is a lack of in-depth exploration into the specic challenges and opportunities oered by ChatGPT, a
state-of-the-art language model, in the academic setting.
The rest of the article is organized as follows. A review of the related literature is presented in Sect.2. Next, in Sect.3,
the methodology adopted in this research is presented. Then, the results are discussed in Sect.4 as the architecture of
ChatGPT (Subsect.4.1), the opportunities (Subsect.4.2), and the challenges (Subsect.4.3) of ChatGPT in the academic
domain. Finally, Sect.5 concludes this article along with future work and limitations.
2 Literature review
Several existing research has focused on outlining the impact of ChatGPT in the academic domain. For example, Lund
and Wang [2] presented dierent contributing concerns of ChatGPT (i.e.: language processing activities, research and
project implementations, ethics and privacy issues, etc. ) in terms of questions and answers. In another article; Curtis
[20] briey explained the strengths (improvement in education, training, clinical care, and research) and limitations (i.e.:
ethical concerns, fake referencing) of this AI model in academic publishing. In another empirical study, AlAfnan etal. [21]
investigated the challenges and opportunities that may arise due to ChatGPT in the case of student-teacher communica-
tion, business writing, and composition courses. After 30 test sessions, it was found that this AI model is used mostly to
seek answers to theory-based questions and generate practical ideas. The challenge was to dierentiate between the
AI-generated text and human written test by the instructors which has a greater chance of deriving incorrect course
outcomes.Meanwhile, Kasneci etal. [22] presented the benets and challenges of LLMs mostly focusing on ChatGPT
in education from a teaching and learning perspectivewhile Cotton etal. [23] found enhanced student engagement,
collaboration, and accessibility as opportunities; and several concerns like academic honesty and plagiarism were raised
Vol.:(0123456789)
Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w Review
due to ChatGPT in the context of academic integrity. In another research,Fuchs [24] discussed the opportunities and
challenges of ChatGPT in higher education. The opportunities included personalized learning, on-demand support, and
independent working ability; and challenges included insucient research data, over-reliance on the bot, biases in the
experimental result, and so on. Sullivan etal. [25] conducted a content analysis on available Western news articles to
derive the impact of ChatGPT on academic integrity and student learning. Whalen and Mouza [26] examined the impact
of AI in K-16 education. ChatGPT can be used to automate the aspects of teaching and learning, be augmented as a
teaching assistant, as a language tool, as a student monitoring tool, and alike.
Two studies conducted by Xames and Shefa [27], and Donmez etal. [28] presented the challenges and opportunities
of ChatGPT in publication and research. Xames found that opportunities like ChatGPT may generate ideas, synthesize
existing works, identify context for the researchers; check article quality, format, plagiarism, and eligibility for the editors;
and evaluate the novelty, quality, clarity, conciseness, coherence, and even the strength and weakness of an article for
the reviewers. On the other hand, Donmez etal. outlined the implication of ChatGPT in nding, checking, the research
question; determining the research design; collecting data; and nalizing the title, etc. with examples. Rahman etal.
[29] highlighted the application of ChatGPT in academic research considering research articles, websites, and visual and
numerical artifacts with practical examples.
From the medical point of view, Alser etal. [30] depicted the involvement of ChatGPT in medical research like author-
ship, plagiarism, and biased result generation. In another research, Homolak [31] derived the opportunities and risks of
ChatGPT in medical, science, and academic publishing. This AI model may perform the tasks of physician and scientist
partially but that is not sucient to replace the involved persons.
Yu [32] presented the stages of the revolution of AI to ChatGPT. He depicted the risks of ChatGPT for the students
with statistics. For example, 89% of college students in America use this AI model to complete their assignments. Also,
teachers in North America are trying hard to motivate students to write articles on their own. But to be frank, these
risks are increasing day by day since it is almost impossible to dierentiate between an AI-generated text and a human
text through reading only. In another research, Quintans etal. [33] outlined some open research questions related to
academia raised due to ChatGPT.
From the related literature review, it can be shown that several existing works have focused on deriving the strengths
and weaknesses of ChatGPT in the context of academic integrity, teaching and learning environment, research and
publication, and the like. None to very few articles combined all these as their research area. So, extensive research with
elaborated and extended visualization of the impact of ChatGPT in academia including education, research, publication,
and academic integrity is required.
3 Methodology
To explore the opportunities and challenges of ChatGPT in academia, the related online resources (news, reports, etc.)
and articles (published in journals, conferences, or preprints) were searched and selected from scholarly databases like
ACM, Scopus, Google Scholar, etc. and Google search engine. Table1 shows a summary of the sources for data collection.
The sources were categorized into seven categories: academic journals, preprint, conference proceedings, books,
technical reports, news articles, blogs, and website articles. The selected literature was then reviewed and analyzed
following the noticing-collecting-thinking model [34]. This data analysis model is generally used for the complex and
rigorous practice of qualitative data analysis to construct a taxonomic structure into three phases—noticing, collecting,
Table 1 Summary of the data
collection source Ser Source type Example Frequency Percentage
1 Academic Journals [14, 36, 37], etc. 17 32.1%
2 Preprints [1, 8, 38], etc. 15 28.3%
3 Conference proceedings [39, 40] 2 3.8%
4 Book [41] 1 1.9%
5 Technical reports [19, 42] 2 3.8%
6 News article [10] 1 1.9%
7 Website and blog [6, 43, 44], etc. 15 28.3%
Total 53 100
Vol:.(1234567890)
Review Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w
and thinking [34, 35]. This structured and comprehensive data analysis approach makes this model creative, resourceful,
and exible. The rst phase encourages one to observe and pay close attention to ne details to develop relevant infor-
mation. Next, in the collecting phase, ndings from the previous phase specically relevant information are gathered
systematically. This includes a more comprehensive analysis of data from a variety range of sources and perspectives to
facilitate the next phase. Finally, information is interpreted intelligently and critically considering dierent viewpoints.
This approach is dierent from other qualitative data analysis approaches since this approach can be adopted in various
contexts like research, problem-solving, and decision-making; and integrating noticing-collecting-thinking a topic or
subject can be well understood before concluding. As such, adopting this approach for this research allowed a recurring
process to reveal the specic concerns (data related to the opportunities and challenges of ChatGPT) from the literature.
Again, for extensive analysis, noticing and following by interpretation must consider any concern as an opportunity or
as a challenge. Thus, the noticing-collecting-thinking model is adopted here for data collection and analysis.
In Noticing phase, the selected literature was meticulously reviewed and tried to notice if any concern is related to
the opportunities and challenges of ChatGPT in the case of academia. In the Collecting phase, all the noticed concerns
(opportunities and challenges) were collected and organized in a tabular format. An example of coding the concerns
is shown in Fig.1. For example, the information "ChatGPT can be used as virtual mentors, voice assistants, innovative
content, smart classrooms, automatic assessments, and personalized learning" (serial 01)—was collected from research
conducted by Shidiq [45]. From this information, the colored portion was marked as codes: i.e. ’virtual mentors, voice
assistants’ and ’personalized learning’ were marked as ’personalized learning’. Thus, codes were extracted from each
information and accumulated for the next phase. All the listed concerns were analyzed to highlight the opportunities
and challenges in Thinking phase. Each concern was labeled as a challenge or opportunity according to its impact on
academia. And nally, a total of 21 opportunities and 16 challenges were found. Then, similar opportunities and chal-
lenges were grouped into four broader themes (research, education, personal skill development, and social). While
grouping revealed concerns (data), a set of criteria was considered that could dierentiate the concerns as similar or
correlated following the Iterative Categorization approach [46]. Also, these criteria could distinguish the categories one
from another. This development of categories organizes data through a variety of dierent divisions. This process was
Fig. 1 Coding for opportunities and challenges from the reviewed literature
Vol.:(0123456789)
Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w Review
iterated several times unless the authors reached a consensus level about which information (opportunities and chal-
lenges) could be considered nal.
Apart from analyzing the related literature and online resources, the ChatGPT application was explored by providing
several questions and the responses generated by ChatGPT were analyzed to reveal its opportunities and challenges in
academia as well.
4 Result andanalysis
This section briey discusses the ndings of the research adopting the noticing-collecting-thinking model. Firstly,
the architecture of ChatGPT is briey described. Then, the opportunities and challenges of this model are presented
respectively.
4.1 ChatGPT architecture
The working process of ChatGPT is depicted in Fig.2 [47]. ChatGPT combines big data, large computing power, and
algorithms to act as an intelligent model [32]. Initially, this model was trained on a massive instruct dataset generated
by OpenAI [48] with numerous parameters [49] through supervised learning [50].
During the Reinforcement Learning from Human Feedback (RLHF) phase, human AI trainers performed conversations
with this model as both a user and an AI assistant. So, this model is able to identify the patterns of input data by deter-
mining the statistical structure within the data [40]. To generation process of the reward model of RLHF was repeated
several times. The conversations in the initial training phase were considered as comparison data. For this, one randomly
selected model-generated message was sampled to some alternative responses and all these samples were ranked by
AI trainers. After that, the model was ne-tuned through these reward models.
Interaction with ChatGPT can be divided into Query and Response. Generally, a user can make a Query to the ChatGPT.
This query is directly processed by the GPT-4. GPT-4 was developed following the transformer architecture [39] which is
highly appropriate for handling long data sequences [36]. The multi-layer transformer network consists of several small
circuitry sets each having a dierent purpose. There are two types of circuitry networks in a transformer architecture.
The rst one is the Feed Forward network which is responsible for extracting information from the token fed into it and
processing the information. Another type of network is Multi-Head Attention circuitry which is designed for Encoder/
Fig. 2 The working process of ChatGPT
Vol:.(1234567890)
Review Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w
Decoder/ Encoder-Decoder self-attention. In the case of encoder and decoder multi-head attention circuitry queries,
keys and values are calculated from the encoder and decoder states respectively. For encode-decoder multi-head atten-
tion, the queries are computed from the decoder state; keys and values are computed from encoder states. The output
generation circuitry generates the probable outputs and then the output data are ne-tuned. After that, ChatGPT is
instructed by GPT-4 to respond. Finally, ChatGPT interacts with a human by providing a human-like response through the
conversational interface. After ChatGPT is publicly available, this model is still being learned through conversations with
global users. As such, all the conversation data are combined with the existing dataset and used for training the GPT-4.
4.2 Opportunities ofChatGPT
In the eld of academia, ChatGPT may have several potential opportunities and applications. These can be divided into
four broad categories (Fig.3):
(a) Research: Since ChatGPT is a highly advanced AI language model, it can make signicant contributions to the
research domain by assisting researchers in various tasks and by facilitating new discoveries. For example,
• Facilitate research activities: ChatGPT can be used as a tool for researchers for example, in the case of NLP, Chat-
GPT may assist in testing and upgrading the state-of-the-art in language understanding, generation, translation,
classication, and categorization, etc [37, 51]. This model is also able to identify dierent parts of speech, entities
(like a person, organization, location, etc.), and sentiment (positive, negative, neutral) for a sentence [51].
ChatGPT can provide up-to-date information on emerging technologies and their applications, making it
easier for individuals to conduct research in these rapidly changing fields. This model can also offer a more
cost-effective and efficient solution in some cases [8]. Figure4 shows a list of articles suggested on a query
for a related research paper to a topic.
Fig. 3 The opportunities of ChatGPT in academia
Vol.:(0123456789)
Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w Review
• Promote creativity and innovation: By automating the repetitive and troublesome tasks, ChatGPT can assist
the researchers as such ChatGPT may assist to collect data, reviewing the literature, and synthesizing data that
helps the researchers to contribute more in creative and higher level tasks [52, 53]. Again, this model can help
with novel and unique ideas/ ways of thinking. An example of this scenario is presented in Fig.5. A question
was asked about how the performance of biometric authentication can be enhanced. ChatGPT suggested
implementing adaptive authentication as a solution.
Compared to other language models, firstly, ChatGPT is trained on larger and more diverse data sets to
adapt a broader range of human language patterns [54]. Second, ChatGPT is more context-aware in case of a
conversation which helps to generate more relevant and accurate responses to a set of context information
[55]. Finally, ChatGPT is still being trained through the queries and responses. So, this model is updated (with
new data and algorithms) to generate innovative responses.
• Improve data analysis and interpretation: The text summarizing capability of ChatGPT can help researchers
with a better understanding of existing knowledge on a specific topic. ChatGPT can analyze large data sets
refined from a large number of texts without human involvement [56] which in turn may produce interesting
patterns that can be difficult to reveal through traditional approaches to data analysis.
Fig. 4 ChatGPT provided a list
of articles related to ’acces-
sibility investigation’
Fig. 5 ChatGPT suggested
innovative ways to enhance
biometric authentication
Vol:.(1234567890)
Review Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w
• Provide additional resources: ChatGPT can provide users with additional relevant information and resources on
a particular topic [52] which may benet in pursuing academic research and projects.
(b) Education and knowledge retrieval: ChatGPT can be used as an educational tool to assist students as well as teachers
in retrieving knowledge and information:
• Enhance collaboration and peer learning: ChatGPT’s sophisticated language processing abilities can facilitate
cross-eld collaboration that may enable students from diverse disciplines to communicate and collaborate
more eectively [22] The model’s ability to answer questions and provide information can encourage knowledge
sharing and peer-to-peer learning among students. ChatGPT’s language generation features can aid in better
communication, promoting the exchange of ideas and knowledge, resulting in more productive collaboration
and peer learning.
• Support needs for special education: ChatGPT has the capacity to address a broad spectrum of disabilities, includ-
ing diculties with reading, writing, and communicating [22] High-quality education for people with disabilities
can be achieved through the implementation of inclusive education practices, the provision of assistive tech-
nology and accommodations, and the training of teachers and sta to support students with disabilities [57].
ChatGPT can allow students to access educational resources like a wide range of subjects, study materials, and
online courses. This can help students from low-income, rural, and remote communities and provide them with
the same educational opportunities as their peers [8].
• Provide personalized tutoring: Since ChatGPT can reply to user questions, it can oer students a personalized
learning experience in their own learning styles and pace [24, 58, 59]. This model is designed to answer the
questions asked of it. If any answer is not understood, this model also supports asking again or regenerating the
answer [60]. Also, ChatGPT is also able to oer resources for the professional development of the teachers [41].
• Build ethics and social responsibilities: ChatGPT allows the user access to information related to ethical and socially
responsible practices. This will help individuals understand the signicance of these practices and how they can
apply them in their personal and professional lives.
• Automate assessment: ChatGPT is able to grade and generate feedback on a given text which can be utilized
as automated scoring and feedback on student assignments and other writings [58]. This will save time for the
teachers and students by providing instant feedback on student assignments [61].
• Improve exam performance: ChatGPT can help students to prepare better for their examinations by exploring a
number of related practice problems and solutions [60].
(c) Personal skill development:
• Improve language skills: ChatGPT can play a signicant role in improving individual language skills. This model
is able to perform machine translation to overcome language barriers and make information more accessible
[38]. Again, this model can improve users’ language prociency by answering questions and providing practice
(speaking and writing) sessions in real time [60].
• Enhance content creation: Since ChatGPT can provide feedback and suggestions on a given text, this feature
can improve individual content creation [44]. ChatGPT can generate written content on a specic topic, helping
content creators save time and eort in writing articles, blog posts, and other written materials [56].
• Provide language translation: ChatGPT can translate text from one language to another, providing real-time
translation services and improving communication between individuals using dierent languages [38]. Moreover,
ChatGPT can aid human translators by oering alternate phrasing, dening uncommon terms, and enhancing
the accuracy of the translation process.
• Improve critical thinking skills: ChatGPT can improve critical thinking skills by providing access to information
and generating new topics as well as debates, encouraging reection and analysis [8]. Individuals can broaden
their knowledge, and evaluate information critically with improved mathematical reasoning and computational
skills, leading to more informed decisions.
• Reduce digital illiteracy: ChatGPT can reduce digital illiteracy by accessing and utilizing technology and provid-
ing individuals with information and support in using it. This helps to increase individuals’ technology skills and
condence, reducing barriers and providing greater opportunities for accessing information and participating
in a tech-driven world [62].
Vol.:(0123456789)
Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w Review
• Develop professional skills: ChatGPT can oer additional resources for a specic topic that can be benecial for
career development and job training [8]. This model can improve professional skills i.e.: project management and
teamwork skills by providing relevant information and support to individuals, helping them to understand project
management methodologies, best practices, and strategies; presentation skills by giving access to information
on eective presentation methods, providing constructive feedback and recommendations [63].
(d) Help and support:
• Provide student service: ChatGPT allows for quicker and more ecient support to students by oering pre-written
responses to questions [64]. This model is able to answer Frequently Asked Questions (FAQ) to support students
and teachers eectively while accessing any online portal related to education [43].
• Oer multi-lingual support: ChatGPT can generate text in multiple languages, making it accessible to a wider
audience across language barriers [38].
• Assist programming projects: ChatGPT is able to provide code snippets and suggestions, generate documenta-
tion and commenting, and help to nd solutions to problems [53].
• Provide 24/7 support: ChatGPT can be used as an online tool making educational resources and support available
to students immediately during not only regular business hours but also at all times [24, 60].
4.3 Challenges ofChatGPT
Though ChatGPT is a powerful tool with numerous potential benefits, some challenges may arise with its use. The
challenges can be categorized into four major categories (Fig.6):
(a) Research:
• Inconsistent performance: Being an AI model, ChatGPT gives machine-generated replies. Though this model is
trained on huge datasets, there is a chance of containing biased data which may generate biased output [24, 60,
Fig. 6 Challenges of ChatGPT in academia
Vol:.(1234567890)
Review Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w
65]. Again, this model may generate dierent replies on the same question when asked separately [62]. Moreo-
ver, ChatGPT may lead to outline inconsistent performance which in turn may aect the reliability and validity of
research ndings [49]. Figure7(a) and (b) shows that ChatGPT oered ve articles related to each topic that was
asked for. For the two ideas that ChatGPT suggested, the authors and years are the same, also the articles don’t
exist.
• Diculty to adopt new models: ChatGPT can be integrated to develop a new system. However, usage and integra-
tion of this model with existing ones can be complex and dicult. Since all queries to this model are stored and
used to train this model further, there is always a data privacy and security concern which may make students
adopt ChatGPT [58].
• Sudden rise of technical issues: ChatGPT is an AI model having limited understanding and background informa-
tion. So, sudden technical issues such as compatibility issues with existing systems, software bugs, or hardware
malfunctions, may occur [58].
(b) Education and knowledge retrieval:
• Provides inadequate support: ChatGPT may provide inadequate user support [1]. If the user doesn’t provide the
clear, concise, and relevant input, ChatGPT may generate an accurate response [24]. This model was developed
following transformer architecture which is highly appropriate for handling text data. So, this model isn’t able
to generate graphs or images. But ChatGPT can provide necessary data and instructions on how to create
a graph using various software tools [66] (Fig.8). This model also can’t provide any direct reference, link, or
source for any responses [66]. This may create difficulties in writing assignments and articles with references
that may ensure the authenticity of the statements. For example, a link for accessing a research article was
Fig. 7 ChatGPT provided ve articleswith same authorsand dierent titlesfor dierent topics (a) articles on biometric authentication meth-
ods(b) articles on biometric generated private/public key cryptography
Fig. 8 ChatGPT is not able to
generate graph
Vol.:(0123456789)
Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w Review
asked to ChatGPT. In response, it was not able to provide an HTML link for the original source of the article
(Fig.9).
• Spread misinformation: ChatGPT is trained on a huge but limited data set up to 2021. In 2023, there is a chance
to provide incorrect information as well as results through this model [49, 60]. Figure10 shows, while ChatGPT
was asked: "When is the next Bonn Climate Change Conference scheduled?"; ChatGPT replied that the schedule
is from 13 to 24 June 2022. This query was made on 17th July 2023. The next Bonn Climate Change Conference
after the date of query is scheduled from November 11 to 22, 2024. This scenario refers that ChatGPT may
generate an inconsistent response.
• Misinterpret student understanding: Incorrect assessment of student understanding may happen if ChatGPT
is not able to properly account for different learning styles or prior knowledge and experiences [67]. However,
it is also difficult for the teachers to differentiate between the AI-generated text and the student’s own writing
which may lead to misinterpreting the actual scenario of student understanding or knowledge [23].
• Lead to update course contents: Since ChatGPT is able to assist students with any textual content, students
may use this model for their assignments, online exams, etc. It is very difficult for a human to detect AI-
generated text. So, the course contents may require to be updated in innovative ways especially assignments
that can’t be solved easily by any AI model [68].
(c) Personal skill development:
• Increase ethical concerns: ChatGPT is able to provide aids related to research and studies. This may increase
ethical concerns for the teachers (i.e.: integrity and plagiarism, accuracy and reliability, privacy and security,
fairness and bias, etc. in responses ) [60]. Again, students may make a wrong usage of ChatGPT to complete
their assignments, online quizzes, etc. pretending that they have done it on their own [23, 32]. These may
affect their ethics as well as personality [69].
• Reduce ability to handle complex scenarios: ChatGPT is operated based on algorithms and pre-existing knowl-
edge, rather than human insight and critical thinking which may reduce compatibility in academia. This can
result in limitations in the ability to cope with complex and new situations or consider multiple perspectives
[42].
• Weaken content creation ability: Since ChatGPT is able to assist both teachers and students with their writing,
regular work, study, and research, some users may become over-reliant on this AI model [24]. This may weaken
individual critical thinking skills reducing productivity to create innovative and new content [45, 70].
Fig. 9 ChatGPT was not able
to provide an HTML link for
an article
Fig. 10 ChatGPT generated
wrong response
Vol:.(1234567890)
Review Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w
(d) Social:
• Reduce reliability: Since ChatGPT can generate inconsistent and incorrect information related to a topic, it may
lead to reduced reliability [1]. This may discourage users from adapting this model for their study and research
purposes.
• Reduce privacy: While using ChatGPT, users may share their sensitive information like corporate business policy,
military tactics, nancial data, or personal data [2]. This data is very private and must not be shared with others
[71, 72]. Since ChatGPT is continuously trained on a vast amount of data (even from the conversation with the
users) to generate a response, sensitive or personal information may be leaked or misused [22].
• Lessen security: ChatGPT can be vulnerable to cyber attacks resulting in theft or loss of data. This can also hamper
the overall system security [73]. Again, ChatGPT is basically trained on the data available over the internet (both
authenticated and unauthenticated). ChatGPT can be manipulated to generate misleading or harmful content,
such as fake news, phishing attempts, or malicious code, which can be used to deceive users or compromise
systems [74].
• Increase unemployment: ChatGPT can automate tasks that were previously performed by human workers (data
entry and processing, language translation, code generation and debugging, content generation) that may result
in reduced workload as well as manpower. In turn, it is also possible to create job replacements for human teach-
ers and tutors [60].
5 Conclusions
ChatGPT is an AI model which is already used by millions of users. This article highlights the possible opportunities and
challenges that could arise due to the innovation of ChatGPT in academia. Although the challenges of ChatGPT include
inaccuracy, fairness, ethical issues, and the like, these can be addressed by improving the eectiveness and responsible
usage of this language model. Again, the conversations with every user are stored and used further to learn the model
(ChatGPT) for improving its accuracy to provide better user support. So, for the time being, the opportunities of ChatGPT
will increase while the problems raised due to ChatGPT will decrease. As such, the implementation of ChatGPT may impact
very positively all concerns related to academia in the near future. Therefore, this article provides an depth understand-
ing of ChatGPT in the academic domain.
By addressing the challenges of ChatGPT depicted in this research, students can easily navigate and utilize emerging
technologies eectively. Another implication of this research can be the identication of the areas to improve further
in ChatGPT by analyzing knowledge gaps, ineective principles, and inecient procedures encountered in academia.
Addressing the areas can assist in the development of more ecient procedures and interfaces. Future studies may focus
on addressing these issues to ensure improved quality education. Performing the usability and UX analysis through
evaluation studies of this AL model will also help to ensure more acceptability among the teachers and students [75].
However, since AI technology is a rapidly growing area of computer science, potential measures must be taken to address
all the challenges that may arise in the future.
This research has a few limitations as well. This study reviewed and analyzed the limited online and published resources
that were available till March 2023. Again, the methodology adopted for analyzing the data is qualitative in nature, which
mostly depends on the analyst or researcher’s skills and experiences. Finally, a few cases were discussed with appropriate
examples and scenarios for better understanding.
Author contributions I.I. and M.N.I. formulated the research idea, I.I. and M.N.I. gathered information from the sources, I.I. categorized into
broader domains, M.N.I. and I.I. nalized the detailed themes, I.I. prepared the draft, M.N.I. supervised the study.
Funding Not applicable.
Declarations
Competing interests The authors declare no potential competing interests.
Vol.:(0123456789)
Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w Review
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adapta-
tion, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article
are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in
the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
References
1. Gilson A., Safranek C, Huang T, Socrates V, Chi L, Taylor RA, Chartash D. How well does ChatGPT do when taking the medical licensing exams?
the implications of large language models for medical education and knowledge assessment. medRxiv. 2022; 2022–12.
2. Lund BD, Wang T. Chatting about chatgpt: how may AI and GPT impact academia and libraries? Library Hi Tech News 40(3), 26–29 (2023)
3. Floridi L, Chiriatti M. Gpt-3: its nature, scope, limits, and consequences. Mind Mach. 2020;30:681–94.
4. Scott K. Microsoft teams up with OpenAI to exclusively license GPT-3 language model. 2023. https://blogs.microsoft.com/blog/2020/09/22/
microsoft-teams-up-with-openai-to-exclusively-license-gpt-3-language-model/. Accessed 10 Jul 2023.
5. AI O. OpenAI. 2023. https://openai.com/api/. Accessed 10 Jul 2023.
6. GPT-4. 2023. https://en.wikipedia.org/wiki/GPT-4. Accessed 27 Jul 2023.
7. GPT-4 is OpenAI’s most advanced system, producing safer and more useful responses. 2023. https://openai.com/product/gpt-4. Accessed 3
Aug 2023.
8. Susnjak T. ChatGPT: the end of online exam integrity? arXiv preprint arXiv: 2212. 09292. 2022.
9. GPT-3. 2023. https://en.wikipedia.org/wiki/GPT-3#:
∼
:text=GPT%2D3%20was%20trained%20on,%2C%20and%20Python%2C%20among%20
others. &text=Since%20GPT%2D3’s%20training%20data,training%20for%20distinct%20language%20tasks. Accessed 10 July 2023.
10. How large language models will transform science, society, and AI. 2023. https://hai.stanford.edu/news/how-large-language-models-will-
transform-science-society-and-ai. Accessed 10 Jul 2023.
11. Devlin J, Chang M-W, Lee K, Toutanova K. BERT: pre-training of deep bidirectional transformers for language understanding. arXiv preprint
arXiv: 1810. 04805. 2019.
12. Dai Z, Yang Z, Yang Y, Carbonell J, Le QV, Salakhutdinov R. Transformer-XL: attentive language models beyond a xed-length context. arXiv
preprint arXiv: 1901. 02860. 2019.
13. Shoeybi M, Patwary M, Puri R, LeGresley P, Casper J, Catanzaro B. Megatron-lm: training multi-billion parameter language models using model
parallelism. arXiv preprint arXiv: 1909. 08053. 2019.
14. Lieber O, Sharir O, Lenz B, Shoham Y. Jurassic-1: technical details and evaluation. White Paper. AI21 Labs 1. 2021.
15. Clark K, Luong M-T, Le QV, Manning CD. Electra: Pre-training text encoders as discriminators rather than generators. arXiv preprint arXiv: 2003.
10555. 2020
16. Rae JW, Borgeaud S, Cai T, Millican K , Homann J, Song F, Aslanides J, Henderson S, Ring R, Young S, etal. Scaling language models: methods,
analysis and insights from training gopher. arXiv preprint arXiv: 2112. 11446. 2021
17. Shen Y, Heacock L, Elias J, Hentel KD, Reig B, Shih G, Moy L. ChatGPT and other large language models are double-edged swords. Radiological
Society of North America. 2023.
18. Bang Y, Cahyawijaya S, Lee N, Dai W, Su D, Wilie B, Lovenia H, Ji Z, Yu T, Chung W, etal. A multitask, multilingual, multimodal evaluation of
ChatGPT on reasoning, hallucination, and interactivity. arXiv preprint arXiv: 2302. 04023. 2023.
19. Korinek A. Language models and cognitive automation for economic research. National Bureau of Economic Research: Technical report; 2023.
20. Curtis N. To ChatGPT or not to ChatGPT? the impact of articial intelligence on academic publishing. Pediatr Infect Dis J. 2023;42(4):275.
21. AlAfnan MA, Dishari S, Jovic M, Lomidze K. ChatGPT as an educational tool: opportunities, challenges, and recommendations for communica-
tion, business writing, and composition courses. J Artif Intell Technol. 2023;3(2):60–8.
22. Kasneci E, Seßler K, Küchemann S, Bannert M, Dementieva D, Fischer F, Gasser U, Groh G, Günnemann S, Hüllermeier E, etal. ChatGPT for good?
on opportunities and challenges of large language models for education. Learn Individ Dier. 2023;103.
23. Cotton DR, Cotton PA, Shipway JR. Chatting and cheating: ensuring academic integrity in the era of chatgpt. Innov Educ Teach Int
2024;61(2):228–39
24. Fuchs K. Exploring the opportunities and challenges of NLP models in higher education: is ChatGPT a blessing or a curse? Front Educ.
2023;8:1166682.
25. Sullivan M, Kelly A, McLaughlan P. ChatGPT in higher education: considerations for academic integrity and student learning. J Appl Learn
Teach. 2023;6(1).
26. Whalen J, Mouza C, etal. ChatGPT: challenges, opportunities, and implications for teacher education. Contemp Issues Technol Teach Educ.
2023;23(1):1–23.
27. Xames MD, Shefa J. ChatGPT for research and publication: opportunities and challenges. Available at SSRN 4381803. 2023.
28. Dönmez İ, Sahin I, Gülen S. Conducting academic research with the AI interface ChatGPT: challenges and opportunities. J STEAM Educ. 2023;
6(2): 101–118.
29. Rahman MM, Terano HJ, Rahman MN, Salamzadeh A, Rahaman MS. ChatGPT and academic research: a review and recommendations based
on practical examples. J Educ Manag Dev Stud. 2023; 3(1): 1–12.
30. Alser M, Waisberg E. Concerns with the usage of ChatGPT in academia and medicine: A viewpoint. Am J Med Open 100036(10.1016) (2023)
31. Homolak J. Opportunities and risks of ChatGPT in medicine, science, and academic publishing: a modern promethean dilemma. Croat Med
J. 2023;64(1):1.
32. Yu H. Reection on whether ChatGPT should be banned by academia from the perspective of education and teaching. Front Psychol.
2023;14:1181712.
33. Quintans-Júnior LJ, Gurgel RQ, Araújo AAdS, Correia D, Martins-Filho PR. ChatGPT: the new panacea of the academic world. Sociedade Brasileira
de Medicina Tropical 56, 0060–2023 (2023)
Vol:.(1234567890)
Review Discover Education (2024) 3:31 | https://doi.org/10.1007/s44217-024-00114-w
34. Seidel J. Qualitative data analyisis. The ethnograph v5 manual, appendix E. Qualis Research 20, 2008 (1998)
35. Islam MN, Islam I, Munim KM, Islam AN. A review on the mobile applications developed for covid-19: an exploratory analysis. IEEE Access.
2020;8:145601–10.
36. Ismail A, Ghorashi NS, Javan R. New horizons: the potential role of openai’s ChatGPT in clinical radiology. J Am Coll Radiol 2023;20(7):696–698.
37. Zaremba A, Demir E. ChatGPT: unlocking the future of NLP in nance. Available at SSRN 4323643. 2023.
38. Jiao W, Wang W, Huang J-t, Wang X, Tu Z. Is ChatGPT a good translator? A preliminary study. arXiv preprint arXiv: 2301. 08745. 2023.
39. Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, Kaise L, Polosukhin I. Attention is all you need. In: Advances in Neural Infor-
mation Processing Systems, vol. 30. Curran Associates, Inc., 2017. https://proceedings.neurips.cc/paper_les/paper/2017/le/3f5ee243547d
ee91fbd053c1c4a845aa-Paper.pdf.
40. Knox WB, Stone P. Augmenting reinforcement learning with human feedback. In: ICML 2011 Workshop on new developments in imitation
learning, vol. 855, p. 3, 2011.
41. Atlas S. ChatGPT for higher education and professional development: a guide to conversational AI. College of Business Faculty and Publica-
tions. 2023.
42. Johnson T. Thoughts on ChatGPT after reading crawford’s why we drive: whatever skill you outsource, atrophies. Technical report 2023.
43. From support chats to a searchable FAQ: how we used ChatGPT. 2023. https://www.100ms.live/blog/chatgpt-support-faq. Accessed 10 July
2023.
44. How businesses can use ChatGPT for content marketing. 2023. https://www.markettailor.io/blog/how-businesses- can-use-chatgpt-for-content-
marketing#::text=Input%20specic%20prompts%20or%20queries,creation%2C%20and%20information%20gathering%20tasks. Accessed
10 July 2023.
45. Shidiq M. The use of articial intelligence-based chat-gpt and its challenges for the world of education; from the viewpoint of the develop-
ment of creative writing skills. In: Proceeding of International Conference on Education, Society and Humanity, vol. 1, pp. 353–357, 2023.
46. Neale J. Iterative categorization (IC): a systematic technique for analysing qualitative data. Addiction. 2016;111(6):1096–106.
47. How does ChatGPT Work? 2023. https://fourweekmba.com/how-does-chatgpt-work/. Accessed 10 July 2023.
48. Aligning language models to follow instructions. 2023. https://openai.com/research/instruction-following. Accessed 31 July 2023.
49. Introducing ChatGPT. 2023. https://openai.com/blog/chatgpt#OpenAI. Accessed 31 July 2023.
50. Dayan P, Sahani M, Deback G. Unsupervised learning. The MIT encyclopedia of the cognitive sciences, 857–859, 1999.
51. Biswas S. ChatGPT and the future of medical writing. Radiology 307(2), 223312 (2023)
52. Dowling M, Lucey B. ChatGPT for (nance) research: the bananarama conjecture. Finance Research Letters 53, 103662 (2023)
53. Haque MU, Dharmadasa I, Sworna ZT, Rajapakse RN, Ahmad H. I think this is the most disruptive technology: exploring sentiments of ChatGPT
early adopters using twitter data. arXiv preprint arXiv: 2212. 05856. 2022.
54. Cai ZG, Haslett DA, Duan X, Wang S, Pickering MJ. Does ChatGPT resemble humans in language use? arXiv preprint arXiv: 2303. 08014. 2023.
55. García-Peñalvo FJ. The perception of articial intelligence in educational contexts after the launch of ChatGPT: disruption or panic? Education
in the Knowledge Society (EKS) 24, 31279 (2023)
56. Guo B, Zhang X, Wang Z, Jiang M, Nie J, Ding Y, Yue J, Wu Y. How close is ChatGPT to human experts? comparison corpus, evaluation, and
detection. arXiv preprint arXiv: 2301. 07597. 2023.
57. Improve accessibility in education with ChatGPT. 2023. https://onedrive.live.com/view.aspx?resid=969B824E4F4E553E!5974
&ithint=le%2cdocx &authkey=!AFxj7opd4-o5o7U. Accessed 10 Jul 2023.
58. Zhai X. ChatGPT user experience: Implications for education. Available at SSRN 4312418. 2022.
59. Munim KM, Islam I, Rahman MM, Islam MN. Adopting HCI and usability for developing industry 4.0 applications: a case study. In: 2020 2nd
International Conference on Sustainable Technologies for Industry 4.0 (STI), pp. 1–6, IEEE, 2020.
60. Qadir J. Engineering education in the era of ChatGPT: promise and pitfalls of generative AI for education. techrxiv.21789434.v1. 2022.
61. Zhai X. ChatGPT and AI: the game changer for education. Available at SSRN. 2023.
62. Krügel S, Ostermaier A, Uhl M. The moral authority of. arXiv preprint arXiv: 2301. 07098. 2023.
63. Meet ChatGPT of Presentations, ChatBCG. 2023. https://analyticsindiamag.com/meet-chatgpt-of-presentations-chatbcg/. Accessed 10 Jul
2023.
64. ChatGPT in customer service: how your business can leverage the technology. 2023. https://www.apizee.com/chatgpt-customer-service/.
Accessed 10 Jul 2023.
65. Dis EA, Bollen J, Zuidema W, Rooij R, Bockting CL. ChatGPT: ve priorities for research. Nature. 2023;614(7947):224–6.
66. Rudolph J, Tan S, Tan S. ChatGPT: bullshit spewer or the end of traditional assessments in higher education? J Appl Learn Teach.
2023;6(1):342–63.
67. What students are saying about ChatGPT. 2023. https://www.nytimes.com/2023/02/02/learning/students-chatgpt.html. Accessed: 10 Jul
2023.
68. Thorp HH. ChatGPT is fun, but not an author. Science 379(6630), 313 (2023)
69. Dehouche N. Plagiarism in the age of massive generative pre-trained transformers (gpt-3). Ethics Sci Environ Polit. 2021;21:17–23.
70. Sallam M, Salim N, Barakat M, Al-Tammemi A. ChatGPT applications in medical, dental, pharmacy, and public health education: a descriptive
study highlighting the advantages and limitations. Narra J. 2023;3(1):103–103.
71. Munim KM, Islam I, Islam MN. A conceptual framework for improving privacy in mobile operating systems. In: 2019 2nd International Confer-
ence on Innovation in Engineering and Technology (ICIET), pp. 1–6, IEEE; 2019.
72. Islam I, Munim KM, Islam MN, Karim MM. A proposed secure mobile money transfer system for sme in bangladesh: an industry 4.0 perspective.
In: 2019 International Conference on Sustainable Technologies for Industry 4.0 (STI), pp. 1–6. IEEE; 2019.
73. Borji A. A categorical archive of ChatGPT failures. arXiv preprint arXiv: 2302. 03494. 2023.
74. Uzun L. Chatgpt and academic integrity concerns: detecting articial intelligence generated content. Lang Educ Technol. 2023;3(1).
75. Munim, KM, Islam, I, Milton, MMR, Ara, LA, Faisal, FA, Islam, MN: Exploring the impact of design technique on usability: A case study on
designing the eHealth websites using card sorting and interactive dialogue model. Eng Rep. 2023;12738.
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional aliations.
