Journal of University Teaching & Learning Practice

Volume 20
Issue 2 Higher education and digital writing in a Article 07
post-pandemic world

2023

Academic Integrity considerations of AI Large Language Models in the

post-pandemic era: ChatGPT and beyond
Mike Perkins
British University Vietnam, Vietnam, mgperkins@gmail.com

Follow this and additional works at: https://ro.uow.edu.au/jutlp

Recommended Citation
Perkins, M. (2023). Academic Integrity considerations of AI Large Language Models in the post-pandemic
era: ChatGPT and beyond. Journal of University Teaching & Learning Practice, 20(2). https://doi.org/
10.53761/1.20.02.07

Research Online is the open access institutional repository for the University of Wollongong. For further information
contact the UOW Library: research-pubs@uow.edu.au
Academic Integrity considerations of AI Large Language Models in the post-
pandemic era: ChatGPT and beyond

Abstract
This paper explores the academic integrity considerations of students’ use of Artificial Intelligence (AI)
tools using Large Language Models (LLMs) such as ChatGPT in formal assessments. We examine the
evolution of these tools, and highlight the potential ways that LLMs can support in the education of
students in digital writing and beyond, including the teaching of writing and composition, the possibilities
of co-creation between humans and AI, supporting EFL learners, and improving Automated Writing
Evaluations (AWE). We describe and demonstrate the potential that these tools have in creating original,
coherent text that can avoid detection by existing technological methods of detection and trained
academic staff alike, demonstrating a major academic integrity concern related to the use of these tools
by students. Analysing the various issues related to academic integrity that LLMs raise for both Higher
Education Institutions (HEIs) and students, we conclude that it is not the student use of any AI tools that
defines whether plagiarism or a breach of academic integrity has occurred, but whether any use is made
clear by the student. Deciding whether any particular use of LLMs by students can be defined as
academic misconduct is determined by the academic integrity policies of any given HEI, which must be
updated to consider how these tools will be used in future educational environments.

Practitioner Notes
1. Students now have easy access to advanced Artificial Intelligence based tools such as
ChatGPT. These tools use Large Language Models (LLMs) and can be used to create
original written content that students may use in their assessments.
2. These tools can be accessed using commercial services built on this software, often
targeted to students as a means of ‘assisting’ students with assessments.
3. The output created by these LLMs is coherent enough for it not to be detected by
academic staff members, or traditional text-matching software used to detect plagiarism,
but falsified references may hint at their use if unchanged by students.
4. The use of these tools may not necessarily be considered as plagiarism if students are
transparent in how they have been used in any submission, however it may be a breach of
academic integrity policies of any given Higher Education Institution (HEI).
5. There are legitimate uses of these tools in supporting the education of students, meaning
HEIs must carefully consider how policies dealing with student use of this software are
created.

Keywords
Artificial Intelligence, Large Language Models, GPT-3, ChatGPT, plagiarism

This article is available in Journal of University Teaching & Learning Practice: https://ro.uow.edu.au/jutlp/vol20/iss2/
07
 Perkins: Academic Integrity considerations of AI Large Language Models in

Introduction
During the COVID-19 pandemic Higher Education Institutions (HEIs) institutions worldwide were
forced to rapidly alter the delivery and assessment of programmes traditionally taught and
assessed in-person, as international restrictions on movement and gatherings prevented
programmes from being delivered as planned (Kaqinari et al., 2021). This rapid transition to online
learning meant that students were faced with entirely new assessment situations, and ever-
changing regulations both from HEIs, and from their respective governments. At the same time,
HEIs were being faced with the challenge of attempting to maintain academic integrity to assure
the quality and standards of their degrees (Clarke et al., 2022; Rapanta et al., 2021) while using
alternative, and often novel, modes of assessment.
Although online learning does not necessarily equate to higher amounts of academic misconduct
occurring amongst students (Grijalva et al., 2006; Stuber-McEwen et al., 2009) online assessment
has been shown to be associated with increased risks to academic integrity (Miller & Young-
Jones, 2012; St-Onge et al., 2022), as well as more cases of academic dishonesty occurring
(Clarke et al., 2022; Lanier, 2006; Watson & Sottile, 2010). The particular situation of the
pandemic has also resulted in a unique set of circumstances which has been demonstrated to
lead to both an increase in detected cases of AD (Henderson et al., 2022; Jenkins et al., 2022;
Lancaster & Cotarlan, 2021), as well as increases in student or academic staff perceptions of AD
(Amzalag et al., 2021; Reedy et al., 2021; Walsh et al., 2021) occurring.
During the period of the pandemic, research has shown that students sought out and adapted to
new technologies (Vargo et al., 2021) as they were faced with large scale disruptions to their
educational experience. As we emerge into a post-pandemic situation of learning, writing, and
assessment, the availability of new digital tools is increasing the options that students have
available to them in supporting assessments involving digital writing. This paper will explore one
of these still-developing technologies that can enable new opportunities in digital writing, but also
raises significant concerns related to academic integrity: Artificial Intelligence (AI) tools using
Large Language Models (LLMs).
This paper delves into the evolution of AI based digital
tools and the emergence of LLMs and discusses Academic Editors
several key areas to better understand LLMs, the key Academic Editors
Section: Special Issue
ethical concerns related to them, and the future of Section:
Senior Editor: Dr Jo-Ann Kelder
their use in digital writing and beyond. Firstly, we Senior Editor: Dr
Guest Editor: [First Last] Johinke
Rebecca
describe and demonstrate the potential that these Associate Editor: [First Last]
Publication
tools have in creating original, coherent text that can Publication
Received: 28 November 2022
avoid traditional methods of detection by text-
Received:20
Revision: [DD MONTH 2022
December YYYY]
matching software. Secondly, we evaluate whether Revision: 31 January 2023
Accepted:
the use of LLM based tools to support students in Accepted: 22 February 2023
Published:
writing assignments can be considered as plagiarism, Published:
Copyright: © by the authors, in its year of first
academic misconduct, or a breach of academic Copyright:
publication.©This
by the authors, is
publication inan
its open
year of first
access
integrity. Thirdly, we identify the potential these tools publication. This publication is an
publication under the Creative Commons open access
have for supporting the education of students, and publication under
Attribution CC the Creative
BY-ND Commons
4.0 license.
Attribution CC BY-ND 4.0 license.
whether academic staff can detect any such use of
1

1
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

these tools. We end with a discussion on how academic integrity policies of HEIs must be updated
to include the recognition of these tools.
By exploring these areas, we contribute to the literature by increasing the awareness amongst
practitioners of an emerging technological tool which may be used by students to evade traditional
methods of detecting breaches of academic integrity. We also identify key considerations for HEIs
as they begin to consider how their academic integrity policies may need to be adjusted to account
for the emergence and use of these tools by students, and highlight specific areas which may
need to be considered. By determining where boundaries may lie in relation to academic integrity,
academic misconduct, and plagiarism, we support HEIs in the broader development of their
academic integrity policies. Therefore, clarifying how student use of these digital tools may be
considered either an acceptable usage of a potentially paradigm shifting educational tool, or an
academic integrity violation.

Literature
The evolution of AI powered digital writing tools
Students have used digital writing tools to aid them in their assessments since their rise to
prominence in the 1980s (Palmquist, 2003), with the development of style and grammar analysis
software integrated into word processor technology. These tools have seen a long history of
evolution and change, and students believe that these tools are able to support them in improving
their writing output (Nobles & Paganucci, 2015). Although digital writing has long been used in
HEIs (Kozma, 1991), new technologies mean that there is a constant need to evaluate which tools
are available to students, and provide clear guidelines on how these may be used.
Apart from the spelling, grammar, and style checkers built into word processing software, tools
have emerged which are designed to go beyond the basics of pointing out more basic errors in
composition and provide additional guidance to students on improving their writing. For example,
Digital Writing Assistants (DWAs) such as Grammarly, WordTune and Perusall, use Artificial
Intelligence (AI) to improve student writing (Fitria, 2021), and are perceived by students to be
helpful tools in improving their work (Cavaleri & Dianati, 2016; ONeill & Russell, 2019). These
tools have also demonstrated a particular ability to support English as a Foreign Language (EFL)
writers in improving their skills at expressing ideas in written English (Gayed et al., 2022; Nazari
et al., 2021; Zhao, 2022), but it is vital that these EFL learners are supported in knowing how to
use these responsibly so that they do not accidentally breach any academic integrity policies.
Another category of AI based software tools which have been used by students to support them
with their writing are those of Automated Paraphrasing Tools (APTs). APTs are applications which
use machine translation in order to transform one text into another (Rogerson & McCarthy, 2017).
These tools were initially developed in order to support commercial enterprises in search engine
optimisation (Q. Zhang et al., 2014) but have been adapted into tools that can be used to support
students in the paraphrasing of text. Although the developers behind these tools may make claims
that paraphrasing using these tools may not constitute plagiarism (QuillBot, n.d.-b), and
technically discourage students from using these tools without correct citation of where original
text was taken from (QuillBot, n.d.-a), these claims are disingenuous, given the wide variety of
policies that may be in place by various HEIs related to what constitutes AD and plagiarism (Sun,

https://ro.uow.edu.au/jutlp/vol20/iss2/07 2
 Perkins: Academic Integrity considerations of AI Large Language Models in

2013). Passing the burden of paraphrasing text to a digital tool and adjusting the output to
enhance readability and reduce the likelihood that any original text will be spotted by text-
matching software is considered to be a case of academic dishonesty and an example of
paraphrasing plagiarism (Roe & Perkins, 2022).
Legitimate uses of this software can be again demonstrated in EFL education, where these tools
can support students being taught the skill of correct paraphrase (Chen et al., 2015; Park & Yang,
2020; Zhao, 2022). However, it is clear from other reports that students do use these APTs for
academically dishonest purposes (Dinneen, 2021), even if this may be unintentional (Prentice &
Kinden, 2018). Spotting whether students have used APTs in their work can potentially be
identified through the ‘word-salad’ that these tools may produce as an output (Rogerson &
McCarthy, 2017), or through technological detection methods which aim to identify machine
translated text (Wahle, Ruas, Foltýnek, et al., 2022; Q. Zhang et al., 2014), but this remains an
emerging challenge, especially as we consider the next evolution of digital writing tools available
to students: LLMs.
The emergence of Large Language Models
Tools using Artificial Intelligence (AI) to support in the creation of original text have emerged in
recent years, in the form of Large Language Models (LLMs) which are able to develop significant
amounts of brand-new text based on short input requests. This contrasts with the APT and DWA
tools described above, which focus primarily on the manipulation of existing text—paraphrasing,
suggesting alterations to text, or providing predictions for sentence completion. These models,
trained on broad data sets are also referred to as foundation models (Bommasani et al., 2022).
These tools started to achieve prominence from 2017 with the emergence of a range of
transformer based machine learning models which allow for improved performance on language
based tasks when compared to previous machine learning techniques (Vaswani et al., 2017). Of
particular note are the BERT transformer developed by Google in 2018, and the subsequent
releases of the Generative Pre-trained Transformer 2 (GPT-2) by OpenAI in November 2019, and
the Generative Pre-trained Transformer 3 (GPT-3) in June 2020 (Dale, 2021). These LLMs have
received significant attention in popular media (Dale, 2021), including multiple reports of the
eventual publication of Alarie and Cockfield’s (2021) journal paper which was generated entirely
by GPT-3. There has also been increasing awareness of these tools and how they relate to the
future of academic and scientific work as demonstrated by the publication in Nature of an overview
of the advances in this area (Hutson, 2022). The broad release to the public in November 2022
of ChatGPT by OpenAI marked a significant increase in the ability of the underlying software to
create new text using further refined models (GPT-3.5), coupled with a markedly improved user
interface (OpenAI, 2022). This release has led to an increased public dialogue on how LLMs may
impact academic integrity, with multiple opinion pieces written on this topic (see for example
Marche (2022) and Hern (2022)), as well as achieving more than 1 million users within five days
of launch (Brockman, 2022).
Integration of LLMs into commercial services
Although LLMs can be accessed directly, there is a relatively steep learning curve in
understanding how to use the software, and some of the more advanced LLMs such as GPT-3
are currently limited to certain geographic areas and require extensive sign-up procedures.
3

3
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

Because of this, an alternative method that students may use to access the LLMs is through
commercial entities who use an Application Programming Interface (API) to connect to the LLMs.
These services are targeted at different groups of customers including copywriters, marketers
and students and allow user-friendly access to the LLMs. Many of these tools integrate features
of DWAs and APTs, and market the product package together as one ‘writing support tool’ or
similar, available as both free and paid for versions.

This inclusion of LLMs into DWAs adds to the potential confusion for students as to whether the
use of these tools would be considered legitimate. A subset of these tools is marketed specifically
to students to support in the development of full length-assessments, which also include options
to generate citations alongside the work. These entities are referred to as Automatic Article
Generators (AAGs) by Abd-Elaal et al. (2022), and are an additional method for students to use
the underlying LLMs on which they are based. Multiple lists exist online which compare these
services against each other, with several of these appearing to be sponsored by some of the
services themselves.
Examples
As an example of the fluency of the output that can be created using these tools, the two
paragraphs of the section above titled: ‘The emergence of Large Language Models’ were used
as an input to ChatGPT. The prompt to the model was: ‘Continue the following text by writing
approximately 300 words explaining what ChatGPT is capable of, including examples of how it
can be used by different users, with a focus on education. Write in a formal academic style, using
in text citations in APA7 style. Provide all references’ The unedited output from this prompt was:
‘ChatGPT-3 is a chatbot that is powered by the GPT-3 language model developed by
OpenAI. It is a large language model that is trained on a broad dataset and is capable
of generating human-like text based on short input requests. The GPT-3 model is
trained using transformer-based machine learning techniques, which have been shown
to outperform previous machine learning techniques on language-based tasks
(Vaswani et al., 2017).
One of the main capabilities of ChatGPT-3 is its ability to generate new text based on
short input requests. This means that it can be used to create original content, such as
articles, blog posts, and even entire papers. This has significant potential for education,
as it could be used to assist students in the writing process, or even to generate
complete papers for assignments.
There are a number of potential ways in which ChatGPT-3 could be used in education.
For example, it could be used as a writing tool to help students generate ideas for
assignments, or to suggest edits and improvements to existing text. It could also be
used as a tool for language learning, as it can be trained to generate text in multiple
languages.
However, the use of ChatGPT-3 in education also raises some ethical concerns. There
are concerns that the use of chatbots like ChatGPT-3 could lead to a decrease in
academic integrity, as students may be tempted to use the tool to generate entire
papers rather than doing the work themselves. There have been several opinion pieces
4

https://ro.uow.edu.au/jutlp/vol20/iss2/07 4
 Perkins: Academic Integrity considerations of AI Large Language Models in

written on this topic (Marche, 2022; Hern, 2022), and it is important for educators and
students to consider the potential consequences of using this technology.
Overall, ChatGPT-3 is a powerful tool that has the potential to revolutionize the way we
approach education. It has the ability to generate original text and assist students in the
writing process, but it is important to consider the potential ethical implications of its
use.
References:
Bommasani, S., Vaswani, A., & Dale, S. (2022). Large Language Models: A Review.
arXiv preprint arXiv:2201.11701.
Dale, S. (2021). The Rise of Large Language Models. Communications of the ACM,
64(6), 44-53.
Hutson, M. (2022). The next step for language models. Nature, 588(7840), 395-397.
Marche, S. (2022). The dangers of AI-generated papers. The Atlantic. Retrieved from
https://www.theatlantic.com/ideas/archive/2022/02/ai-generated-papers-threaten-
academic-integrity/617370/
Hern, A. (2022). Chatbots could undermine academic integrity, warn researchers. The
Guardian. Retrieved from
https://www.theguardian.com/education/2022/mar/01/chatbots-could-undermine-
academic-integrity-warn-researchers
OpenAI (2022). ChatGPT: A chatbot powered by GPT-3. Retrieved from
https://openai.com/blog/chatgpt/
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., &
Polosukhin, I. (2017). Attention is all you need. In Advances in neural information
processing systems (pp. 5998-6008).’
(OpenAI, n.d.)
This output was created in approximately 5 seconds at zero cost (ChatGPT is currently in a free
research preview stage and costs are likely to change in the near future). The output retained the
language style used in the input and expanded on the discussions by providing examples
specifically related to education as requested. Although in text citations and references were
provided (including references from the input text), these were all fabricated, with titles, URLs,
and journal names being incorrect. However, given that the authors and dates provided in the in-
text citations are correct, this may go unnoticed unless thoroughly checked. This does open a
potential avenue of detection by academic staff if there is a suspicion that these tools may have
been used by students.
A major benefit of ChatGPT as opposed to previous versions of the GPT-3 software is that of the
iteration possible through natural language ‘discussions’ held with the software. For example, in
the same conversational ‘thread’ as the previous output, ChatGPT was asked to: Continue your
response, focusing on how Higher Education Institutions may need to adjust their academic
integrity policies based on student use of these technologies. Again, use full APA 7 in text citations
and provide references. The unedited output created was:

5
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

‘As the use of ChatGPT-3 and other large language models becomes more
widespread, it is likely that Higher Education Institutions (HEIs) will need to
adjust their academic integrity policies to take into account the potential use of
these technologies by students.
One of the main concerns with the use of chatbots like ChatGPT-3 is the
potential for students to use them to generate entire papers or assignments,
rather than doing the work themselves (Marche, 2022). This could lead to a
decrease in academic integrity, as students may be able to pass off the work of
the chatbot as their own. To address this concern, HEIs may need to update
their policies to specifically prohibit the use of chatbots or other AI-powered tools
for the generation of academic work.
In addition to updating policies, HEIs may also need to invest in tools and
technologies to detect the use of chatbots in academic work. Plagiarism
detection software, such as Turnitin, can be used to identify text that has been
generated by chatbots or other AI tools, but these tools may need to be updated
to be able to accurately detect the use of chatbots (Hern, 2022).
Another potential issue with the use of chatbots in education is the potential for
the chatbot to produce content that is biased or offensive. Chatbots are trained
on large datasets, and if these datasets contain biased or offensive content, the
chatbot may produce similar content (Bommasani et al., 2022). To address this
concern, HEIs may need to consider implementing policies or guidelines related
to the use of chatbots or other AI tools, and providing training to students on the
appropriate use of these technologies.
Overall, the use of chatbots like ChatGPT-3 in education raises a number of
ethical concerns that HEIs will need to consider when developing or updating
their academic integrity policies. It is important for HEIs to carefully consider the
potential consequences of the use of these technologies, and to implement
policies and guidelines that ensure the integrity of academic work.
References:
Bommasani, S., Vaswani, A., & Dale, S. (2022). Large Language Models: A
Review. arXiv preprint arXiv:2201.11701.
Marche, S. (2022). The dangers of AI-generated papers. The Atlantic.
Retrieved from https://www.theatlantic.com/ideas/archive/2022/02/ai-
generated-papers-threaten-academic-integrity/617370/
Hern, A. (2022). Chatbots could undermine academic integrity, warn
researchers. The Guardian. Retrieved from
https://www.theguardian.com/education/2022/mar/01/chatbots-could-
undermine-academic-integrity-warn-researchers’
(OpenAI, n.d.)
Again, the references were falsified and there were some inaccuracies related to the claimed
ability of Turnitin to detect AI generated content, but this fluent, five-paragraph output is typical of
6

https://ro.uow.edu.au/jutlp/vol20/iss2/07 6
 Perkins: Academic Integrity considerations of AI Large Language Models in

a response by ChatGPT. It is evident that with some manipulation of the text through iteration
using the tool, along with some minor editing or adjustments of references, a student intending to
create large blocks of text to include in an assessment could rapidly complete certain tasks with
limited effort and costs.
Limitations and newer LLMs
As demonstrated above, the output produced by the LLMs is fluent, however there are some
limitations to the currently available LLMs. Given that the development of these LLMs require pre-
training before they are released, they are unable to answer questions or complete tasks which
relate to very recent events. As they rely on the content they have been trained with to generate
their output, there is the potential for LLMs to produce text that lacks semantic coherence (Dale,
2021), as well as lexical diversity (Gehrmann et al., 2019) through the repetition of expressions
(Dehouche, 2021; Fröhling & Zubiaga, 2021).
Although one concern levied against LLMs is that they will generally always provide an answer or
output, even if that is factually incorrect or misinterpreted (Dale, 2021; Hutson, 2022), recent
research has shown that GPT-3 can be trained to state how confident it is in producing an answer
that is factually correct (Lin et al., 2022), which can be used to support users of the LLMs in their
use of the tools. ChatGPT in particular can also be challenged if the user believes there are
incorrect elements in the output, and can adjust its response accordingly based on any corrections
provided by the user.
Recent concerns related to the lack of transparency regarding the models used in currently
available LLMs, as well as their computational cost and associated environmental impacts (Sanh
et al., 2022; S. Zhang et al., 2022) have resulted in the development of alternative LLMs, including
an open source LLM called GPT-J by Narrativa, and a new suite of Open Pre-trained
Transformers (OPTs) by Meta. The resulting OPT-175B model by Meta has been shown to have
comparable results to GPT-3 in terms of text creation ability, whilst being more open to scrutiny
by researchers (S. Zhang et al., 2022) as to the ethical implications of these models; a particularly
relevant concern as it relates to the study of academic integrity.

Academic integrity considerations of LLMs

Concerns for academic integrity
As demonstrated by the examples provided above, the current generation of LLMs are already
fluent in their output, and emerging research has suggested that existing LLMs can produce
output which humans struggle to identify as being machine created (Abd-Elaal et al., 2022; Clark
et al., 2021; Gunser et al., 2021; Köbis & Mossink, 2021; Wahle et al., 2021; Wahle, Ruas,
Kirstein, et al., 2022). At the same time, given that the text that is produced by LLMs is uniquely
created based on the inputs provided, current research suggests that use of the created text by
students is unlikely to be spotted by existing text-matching software tools used by HEIs (Wahle,
Ruas, Kirstein, et al., 2022). LLMs therefore represent a clear potential threat to academic integrity
as academic staff may be unable to correctly identify the amount of content produced by a student,
and therefore provide an accurate evaluation of a student's comprehension and interpretation of
the topic at hand.

7
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

Definitions
There is a growing body of academic research which is exploring the concerns related to the use
of LLMs in student work (Dehouche, 2021; Eaton et al., 2021; Kumar, Mindzak, Eaton, et al.,
2022; Kumar, Mindzak, & Racz, 2022; Wilder et al., 2021). A question which has been raised by
several of these authors but has yet to be fully answered is whether the use of LLMs by students
can be considered a breach of academic integrity or not. Reaching this determination is
challenging as it is dependent on how various HEIs define terms related to academic integrity.
For this paper, we use the definition of academic integrity provided by the Tertiary Education
Quality and Standards Agency (TEQSA) of ‘the expectation that teachers, students, researchers
and all members of the academic community act with: honesty, trust, fairness, respect and
responsibility’ (Tertiary Education Quality and Standards Agency [TEQSA], 2020).
Recognising the broadness of such a definition, we regard academic misconduct as a breach of
academic integrity principles as specified in policy by a HEI, whether this intentional or not. This
clarification is needed given the differences between what HEIs and students might classify as
academically dishonest behaviour (Parkinson et al., 2022) and as students can unintentionally
breach these policies (Amigud, 2020). A key area of interest of related to the use of LLMs is the
act of plagiarism which we define as ‘misrepresenting the effort that has been carried out by the
author of a written document’ (Perkins et al., 2019, p. 5). Although there are existing and well
known technological tools to detect simple copy and paste plagiarism by students, we are now
operating in an era of more complex ‘second generation’ (Malesky et al., 2016) plagiarism, which
requires ways of thinking to consider what plagiarism really entails, as well as more complex
method of detection.
Student uses of LLMs
Based on these definitions we need to consider a range of possible ways that students may use
an LLM to support them in the writing of an assessment. Doing this helps us to understand where
we may draw a line between acceptable practice, breaches of academic integrity, academic
misconduct and/or plagiarism.
Consider a situation where a student uses an LLM to support them in their work, includes directly
copied output of the tools in their submission, and does not state that they used an AI tool. This
is similar to a case reported by Dinneen (2021) in reference to student use of an APT, where the
student in question was convinced that because they had altered the text of the original author,
even through the use of a software tool, this was not plagiarism. However, based on the definitions
provided above, given the misrepresentation that have been made, usage of this type could be
seen as a relatively clear case of plagiarism by the student (Roe & Perkins, 2022). Student usage
of an LLM or LLM based tool to create an entire essay, report, or other assignment without stating
where this has come from, or one that is presented in a way that was trying to mislead the reader
through false citations, would be even more clearly a case of plagiarism, and likely also a broader
breach of academic integrity principles.
However, the picture becomes less clear if the use of LLMs in student writing has been clearly
stated (Kumar, Mindzak, Eaton, et al., 2022), if any output text has been modified, edited or
enhanced, or if output text has been used as a starting point for the formation of an argument. If
additional scholarly work has been carried out by the student, such as the integration of the output
text with the use of examples and cited sources, then this complicates the situation further. If the
work carried out by a student hasn’t been misrepresented, then it does not fall under the strict

https://ro.uow.edu.au/jutlp/vol20/iss2/07 8
 Perkins: Academic Integrity considerations of AI Large Language Models in

definition of plagiarism, but it may be considered as academic misconduct depending on the

specific policies and regulations in place at individual HEIs. HEIs must therefore be clear in both
their policies and practices in clarifying how these tools may be used by students. The difficulty in
establishing these policies is even more challenging if we consider the use of LLMs as a tool for
student cognitive offloading.
Cognitive offloading
If student use of LLMs could be considered to be academic misconduct if their use isn't clearly
identified, to what extent should we expect a student to state whether they have used LLMs in the
creation of their work? It would be unrealistic to expect a student to identify every sentence which
has been modified using a spelling or grammar checker, every reference which has been
generated by a reference manager, or even any sentences which were completed by the built-in
features of DWAs or Microsoft Word’s write-ahead function. As Microsoft have signed an
exclusive agreement with OpenAi to use GPT-3 and integrate these into Microsoft’s Azure
platform (Langston, 2021), it is highly likely that LLMs will be integrated more deeply into suites
of products that are often provided to students by HEIs in the future (Dale & Viethen, 2021).
Therefore, we may consider LLMs as another potential tool for reducing the cognitive demands
required by a task, described by Risko and Gilbert (2016) as a process called cognitive offloading.
Dawson (2020) has described how assessment tasks can be developed in a way which makes
the use of any such cognitive offloading tools transparent, and a specific usage of LLMs as an
example of this is discussed by Dawson (2022, as cited in Sparrow, 2022).
Where then, should this distinction be drawn? Given that many of the more advanced DWAs have
multiple features available to students which already claim to integrate AI technologies, it is
feasible to believe that a student could very easily and inadvertently breach any policies related
to their ‘correct’ usage, especially when LLMs become even more deeply integrated into DWAs.
This already appears to be the case in the example of one DWA called Write Full: a full-service
DWA which has integrated paraphrasing tools which may encourage students to use the work of
others as inputs, rather than their own writing. Adding LLM based features to the next generation
of DWAs to suggest additional points for students to consider, and even offering to write
‘suggested’ text is a very feasible next step for developers wishing to retain an edge over their
competitors in this emerging market.
Plagiarism or Academic Misconduct?
Overall, it is highly challenging to unpick the complexities surrounding student usage of LLMs and
tools integrating these, and to make ethical judgements as to whether their use may be considered
acceptable or not. Kumar (2022) examines how the use of LLMs may be considered from a range
of ethical perspectives, and the broader ethical concerns related to LLMs are considered by Luitse
and Denkena (2021), but critical questions remain as to how their use may be encouraged or
discouraged on a policy level by HEIs.
Based on the analysis above, we conclude that the use of LLMs or LLM based tools should not
be considered as plagiarism or a breach of academic integrity if how these tools have been used
is stated clearly and made transparent by students. However, given the significantly more
advanced capability of LLM based tools to create new content (as opposed to the refinement of
existing content) HEIs must be aware of the potential danger of learning outcomes not being met
if LLMs are being used to develop content, rather than this being created by students themselves.

9
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

LLM use cases in the classroom: Digital writing and beyond

As discussed above, there are significant academic integrity concerns related to how HEIs must
adapt to the potential use of these AI tools in student submissions. However, research has
highlighted several legitimate use cases of AI technologies integrated into digital tools which can
support in the education of students. These include specific use cases by students in writing and
composition classes and in the development of creative outputs, and also how academic staff
may integrate these tools in the areas of supporting EFL learners and improving Automated
Writing Evaluation (AWE).
Writing and composition
McKnight (2021) highlights the specific challenges of integrating AI into the teaching of writing
and composition, and proposes that composition teachers working with AI should guide students
in:
‘..when it is appropriate to incorporate AI in writing projects, what the affordances and
constraints of different versions of AI might be, whether and how to acknowledge the role
of AI, and what dimensions humans can contribute beyond the efficiency of machines.’
(McKnight, 2021, p. 11).
Using LLMs in the teaching of digital writing therefore becomes a method to support writers in
engaging more deeply with a topic, rather than at a surface level. McKnight (2021) discusses how
the integration of LLMs into student writing has the potential to encourage new modes of thinking,
but that this must be done with consideration as to how AI writing might be used in the future
workplace.
Writing inputs and adjusting model parameters for LLMs in a way that results in legible, usable
text is a specific technical skill which could also be integrated into writing instruction sessions.
Doing so would allow students to improve their meta-linguistic knowledge around a topic (Godwin-
Jones, 2022) and therefore improve their potential skills in writing in specific areas.
McKnight (2021) and Anson (2022) both recognise the importance of training students in how
LLMS be used as a form of writing co-creation, as well as discussing when this may or may not
be appropriate. Taking this approach would require the development of clear policies on both a
programme and institutional level to clarify to students exactly how this may be done in student
work, and the level of reporting required to avoid any breaches of academic misconduct policies.
Limits regarding how much of the end output must be the student’s own work as opposed to that
of an AI tool could also be provided to help support students in how to use these tools in an
acceptable way.
However, there have already been concerns raised from writers who believe that co-writing with
the current generation of LLMs presents significant barriers regarding the control of the writing
process (Biermann et al., 2022), and this highlights the further explorations required in the field of
human-AI co-creation to better understand the future needs of the writing classroom.
Creative outputs
LLMs have a demonstrated capability in not only traditional writing activities, but also for creative
works such as poetry (Gunser et al., 2021; Köbis & Mossink, 2021) and computer programming
(Biderman & Raff, 2022). Other tools also produced by OpenAI include the Dall·E 2 AI system

10

https://ro.uow.edu.au/jutlp/vol20/iss2/07 10
 Perkins: Academic Integrity considerations of AI Large Language Models in

which is able to create and edit detailed digital artwork based on natural language prompts
provided by users (OpenAi, n.d.)
As an increasing amount of textual and creative outputs are being co-created by humans and AI
based technologies, there is ongoing research exploring how the intricacies of co-creation may
be further be codified and explored in the fields of creative writing (M. Lee et al., 2022; Y. Lee et
al., 2022), software development (Biderman & Raff, 2022) and artistic outputs (Oppenlaender,
2022). Any such co-creation also raises concerns related to copyright (Dehouche, 2021) and how
these systems can be more transparent, fair, and accountable (Fröhling & Zubiaga, 2021;
Oppenlaender, 2022).
Use cases by academic staff to support students
Deciding how to identify what is acceptable use of these tools in relation to student submissions
of any writing or creative outputs will be a major challenge for HEIs. Using the definitions related
to plagiarism and academic integrity discussed above, if the use of AI tools is made clear in any
submission of work by a student, the use cases above could not be considered as plagiarism.
However, this does highlight again how HEIs must ensure that any academic integrity policies are
explicit in how these tools may be used, as opposed to a simple blanket ban on their use,
especially considering the broader benefits that these tools may provide in the areas of supporting
EFL learners and in AWE.
Supporting EFL learners
HEIs must also consider how AI based tools are used in language training for EFL students before
they enter their main academic programmes. Chen et al., (2015) describe the use of a corpus
based software program similar to an APT to support Chinese students in their study of English.
Although this tool does not claim to use AI technology, it introduces students to a digital method
which may be used through trial and error to create better textual outputs in their paraphrasing.
Gayed et al., (2022) demonstrate how an LLM tool based on GPT-2 can be helpful for EFL
students writing English by suggesting predictions which expands on the more basic support
offered by existing writing tools. Their study suggested that this tool could support in improving
the lexical diversity of student work, but statistically significant results were not obtained.
These studies highlight the specific need for education of students on any policies which
discourage the use of these tools on core academic programmes. If students are trained on these
tools in pre-university language programmes, and then move into their main studies, it is easy to
see that students would expect that they would be able to continue using them. This is especially
important given the increased risks of academic misconduct occurring in students with lower
English language abilities (Perkins et al., 2018). Therefore, it is clear that academic integrity
policies needs to be context sensitive (Price, 2002) as students transition throughout their studies
and recognise the needs of different groups of students, especially ‘international’ students (Fatemi
& Saito, 2020).
Automated Writing Evaluation (AWE)
Using software tools to support in AWE of written text and providing feedback to students has
been demonstrated to be of benefit to writing instructors and students alike (Fu et al., 2022; Strobl
et al., 2019; Warschauer & Grimes, 2008). The further development of LLMs may improve this,
especially if this integrated into custom tools to support in AWE (Lim et al., 2022), or DWAs such
as Grammarly (Godwin-Jones, 2022). Although Grammarly uses its own AI supported tool to
correct mistakes in grammar and rather than relying on existing LLM transformers (Grammarly,
11

11
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

2022) the use of LLMs in these tools can particularly reduce the burden on EFL instructors
(Godwin-Jones, 2022).
Some of the recognised shortcomings of present AWE tools are the generic nature of the
feedback provided (Fu et al., 2022) and an inability to focus feedback on how well a task has
been addressed (Allen et al., 2016). The new capabilities of ChatGPT to provide effective critique
on creative writing, equivalent to that of human reviewers (Furze, 2022), suggests that these
shortcomings may be addressed by LLM based tools.

Can the use of LLMs be detected by academic staff?

Any actions that HEIs could choose to take against students who use these tools in an
unacceptable manner can only be carried out if the outputs of these tools can be identified as
such by academic staff responsible for the evaluation of student work. We review the current
evidence evaluating how possible this is.
GPT-2 studies
Abd-Elaal et al., (2022) present a study highlighting the difficulty that academic staff may have in
identifying output produced by LLMs, and the role that training may play in supporting in the
correct identification of the usage of any such tools. Their results indicate that on average,
participants were able to correctly identify sample text as generated by either a human or an LLM
at a rate of 59.5%, barely higher than chance alone. Although it is encouraging to note that training
of academic staff resulted in an increased ability of academic staff to correctly identify the text
sections, the samples were generated using GPT-2 as opposed to GPT-3.
A similar lack of identification ability has been found in creative outputs. Köbis & Mossink (2021)
found that GPT-2 created poems could not be reliably identified as such when the authors
selected the best outputs created by an LLM, and Gunser et al. (2021) found that poem
continuations could not always be correctly identified by professionals with a literature-specific
background as machine or human written.
GPT-3 studies
Studies assessing the ability of GPT-3 produced output show that as more complex LLMs are
used, the ability of humans to detect material drops even further. Kumar et al. (2022) presents
preliminary work aimed at identifying whether study participants drawn from a range of
backgrounds were able to identify whether text was developed by humans or by GPT-3. They
found that regardless of background, participants found this to be a challenging task, with a high
likelihood of ascribing the AI writing samples to humans. Although this study was not focused on
how academic staff may identify machine created output, it highlights the challenges present in
both the academy and beyond when it comes to identifying authorship of text.
Clark et al. (2021) also assessed the ability of non-expert evaluators to identify whether text was
produced by a human, or a LLM (GPT-2 and GPT-3), but used a large sample (n=780) accessed
through the Amazon Mechanical Turk platform. Their results showed that evaluators were able to
identify GPT-2 produced text at an accuracy rate of 57.9%, but GPT-3 at a rate of only 49.9%.
Providing training to participants using examples of LLM generated text (specifically to
demonstrate the ability of LLMs to produce ‘creative’ output) marginally increased the ability of
participants to correctly identify text as machine created. These studies highlight the rapid

12

https://ro.uow.edu.au/jutlp/vol20/iss2/07 12
 Perkins: Academic Integrity considerations of AI Large Language Models in

improvements of LLMs to create text that cannot be easily identified by study participants as
machine created.
Methodological concerns
An unavoidable methodological issue with these experimental studies is that in order to determine
whether participants can accurately identify text as machine or human generated, participants
need to be aware that some of the text they are about to encounter may be machine generated
before participating in an experiment. Given the novelty of these tools, it is likely that even
experienced academic staff are simply not aware of the capabilities that these tools have, as
demonstrated with the participants in Clark et al.’s study (2021). This may result in their ability to
identify any LLM produced output ‘in situ’ when evaluating work being even lower than
demonstrated in an experimental design.
However, given the limited amount of empirical evidence available, and the potential significant
threats that LLMs pose to academic integrity, further research is needed in this area to support
HEIs in understanding how academic staff may be trained or supported to detect the use of AI
tools in student work.
Technological methods of detection
Technological methods to identify text generated by LLMs have been proposed by Gehrmann et
al. (2019) who present the results of a study using a tool named GLTR. This study demonstrates
that using GLTR can improve the ability of participants to correctly identify detection of machine
generated content from 54% to 74%. However, this tool was tested against GPT-2 produced
output and used students rather than academic staff as their participants. Solaiman et al. (2019)
and Ippolito et al.(2020) also both present tools which showed an encouraging ability in detecting
machine created text, although these tools were not tested against the latest generation of LLMs.
Fröhling & Zubiaga (2021) present a promising low cost detection model which is able to
accurately detect machine created text created using GPT-2 and GPT-3, but highlight the ethical
challenges of deploying any such detectors which may potentially discriminate against EFL
students by incorrectly identifying human created text as machine written —a particular concern
in HEIs with a high concentration of non-native English speaking students. While these results do
show promise for the potential future possibilities of software to support in the identification of
LLM produced output, Bidermann and Raff (2022) have already demonstrated how more
advanced models (GPT-J) can fool machine detection programmes such as MOSS. This suggest
that tools using more advanced LLMs may be even less detectable by technological means.
Following the increased interest from the general public and academics alike in AI assisted writing
after the release of ChatGPT in November 2022, individuals and organisations have either
released, or have announced the imminent release of tools which claim to have the ability to
detect AI generated text. These tools include GPTZero (https://gptzero.me/) and Crossplag AI
detect (https://crossplag.com/ai-content-detector/) and do show promise in being able to detect
the use of AI generated text. However, further study is required to identify the accuracy of these
tools, as well as their suitability for use in academic settings to avoid inadvertently accusing
students of potential breaches of academic misconduct. Any tools used to support in the machine
detection of LLM output must be continually re-evaluated as new LLMs emerge, as well as
methods to avoid detection of any tools are developed, resulting in an ongoing ‘arms race’
scenario (Roe & Perkins, 2022).

13

13
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

Given that both academic staff, as well as technological methods of detection are unable to
accurately detect machine generated text and therefore student uses of LLM based tools, this
presents a clear threat to academic integrity for HEIs, requiring a range of adjustments to be made
in both practice and policy.

Academic integrity policy adjustments

Although a certain amount of cognitive offloading using DWAs and other digital tools may be
considered acceptable by HEIs, there needs to be careful development of the academic integrity
policies of HEIs to clarify how LLMs based tools may be used by students, and whether their use
may be considered as academic misconduct.
Academic integrity policies must recognise these tools by name to make it clear to students and
staff that the HEI is aware of this technology and how it may be used by students and staff. Any
limits to their use, or statements which students should provide if these tools are used should be
stated and unambiguous. Specific examples of acceptable and unacceptable usage should be
provided to improve the understanding of students, as well as staff involved in any academic
misconduct decisions. The final policy should also be widely communicated to students and staff,
with training given to support whichever approach is taken.
Given that existing research has demonstrated that academic staff are unlikely to be able to
accurately identify the usage of LLMs in student work, a blanket ban of these tools is likely
unenforceable. This lack of enforceability, coupled with the clear benefits that these tools may
bring to the education of certain groups of students, means that an approach which seeks to
completely ban the usage of LLM based tools is therefore not recommended. A policy approach
which may be more suitable could instead have a more nuanced approach which recognises the
potential benefits that LLMs may bring, the evolving social understanding of plagiarism, and the
changing nature of digital writing and human-AI co-creation.

Conclusion
What we as academic staff define as plagiarism is rapidly changing due to the social construction
of this term (Anson, 2022). Although evidence suggests that plagiarism overall has decreased
between 1990 and 2020 (Curtis, 2022), the COVID-19 pandemic has caused an increase in both
the detected number of academic dishonesty cases (Henderson et al., 2022; Jenkins et al., 2022;
Lancaster & Cotarlan, 2021), as well as increases in student or academic staff perceptions of
academic dishonesty occurring (Amzalag et al., 2021; Reedy et al., 2021; Walsh et al., 2021).
This paper has explored a rapidly developing new category of digital tools which are of concern if
we wish to maintain academic integrity in a post-pandemic world: AI based Large Language
Models and associated tools and software.
We have identified how LLMs can produce coherent, original text that students may potentially
use in assessments, and conclude that LLMs have already progressed to the point that neither
trained academic staff or technological tools can consistently determine whether text is generated
by an LLM or by a human.
Deciding whether any particular use of LLMs by students may be defined as academic misconduct
will be determined by the future policies of any given HEI, and this highlights the importance of
creating clear academic integrity policies and educating students in any acceptable use cases of
LLMs. Although we have identified the potential ways that LLMs can support in digital writing and

14

https://ro.uow.edu.au/jutlp/vol20/iss2/07 14
 Perkins: Academic Integrity considerations of AI Large Language Models in

beyond, the inability to accurately detect whether LLMs have been used presents a clear threat
to the academic integrity of HEIs. However, it is not the use of the tools themselves that defines
whether plagiarism or a breach of academic integrity has occurred, but whether any such use is
made clear.
Although it is unlikely that AI will ever take over the traditional role of a teacher (Cope et al., 2021),
the future development of LLMs and broader AI supported digital tools have a strong potential for
improving the experiences of students and teachers alike in the next generation of HEI
classrooms, both in writing instruction and beyond. This view is echoed by Godwin-Jones (2022)
who highlight the possible co-creation role that these systems may have to play in future
educational scenarios. It is clear from the studies presented in this paper that there are significant
challenges still left to address in this rapidly developing area. Given that the use of the current
generation of LLMs cannot be accurately detected by academic staff or technical means of
detection, the likelihood of accurately detecting any usage of these tools by students in their
submissions typical academic will likely not improve and may even decrease further as new LLMs
are developed.
This situation, coupled with the identified difficulties in determining whether we can even consider
the use of such tools to be a breach of academic integrity, and the potential benefits of LLM based
tools means that a blanket ban of these tools at an institutional level is neither feasible, nor
enforceable. Despite the potential threats to academic integrity presented in the paper, we believe
that the future integration of LLMs and other AI supported digital tools into the classroom
environment is highly likely, and therefore HEIs must consider the implications of this in future
policy development.

15

15
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

References
Abd-Elaal, E.-S., Gamage, S. H., & Mills, J. E. (2022). Assisting academics to identify computer
generated writing. European Journal of Engineering Education, 1–21.
https://doi.org/10.1080/03043797.2022.2046709

Alarie, B., & Cockfield, A. (2021). Will machines replace us?: Machine-authored texts and the
future of scholarship. Law, Technology and Humans, 3(2), 5–11.
https://doi.org/10.5204/lthj.2089.

Allen, L. K., Jacovina, M. E., & McNamara, D. S. (2016). Computer-Based Writing Instruction. In
Handbook of writing research (Second edition, pp. 316–329). Guildford.

Amigud, A. (2020). Cheaters on Twitter: An analysis of engagement approaches of contract

cheating services. Studies in Higher Education, 45(3), 692–705.
https://doi.org/10.1080/03075079.2018.1564258

Amzalag, M., Shapira, N., & Dolev, N. (2021). Two Sides of the Coin: Lack of Academic Integrity
in Exams During the Corona Pandemic, Students’ and Lecturers’ Perceptions. Journal of
Academic Ethics. https://doi.org/10.1007/s10805-021-09413-5

Anson, C. M. (2022). AI-Based Text Generation and the Social Construction of “Fraudulent
Authorship”: A Revisitation. Composition Studies, 50, 37–46.
https://compositionstudiesjournal.files.wordpress.com/2022/07/anson.pdf

Biderman, S., & Raff, E. (2022). Fooling MOSS Detection with Pretrained Language Models
(arXiv:2201.07406). arXiv. https://doi.org/10.48550/arXiv.2201.07406

Biermann, O. C., Ma, N. F., & Yoon, D. (2022). From Tool to Companion: Storywriters Want AI
Writers to Respect Their Personal Values and Writing Strategies. Designing Interactive
Systems Conference, 1209–1227. https://doi.org/10.1145/3532106.3533506

Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S.,
Bohg, J., Bosselut, A., Brunskill, E., Brynjolfsson, E., Buch, S., Card, D., Castellon, R.,
Chatterji, N., Chen, A., Creel, K., Davis, J. Q., Demszky, D., … Liang, P. (2022). On the
Opportunities and Risks of Foundation Models (arXiv:2108.07258). arXiv.
http://arxiv.org/abs/2108.07258

Brockman, G. [@gdb]. (2022, December 5). ChatGPT just crossed 1 million users; it’s been 5
days since launch. [Tweet]. Twitter.
https://twitter.com/gdb/status/1599683104142430208

Cavaleri, M. R., & Dianati, S. (2016). You want me to check your grammar again? The
usefulness of an online grammar checker as perceived by students. Journal of Academic
Language and Learning, 10(1), Article 1.
https://journal.aall.org.au/index.php/jall/article/view/393

16

https://ro.uow.edu.au/jutlp/vol20/iss2/07 16
 Perkins: Academic Integrity considerations of AI Large Language Models in

Chen, M.-H., Huang, S.-T., Chang, J. S., & Liou, H.-C. (2015). Developing a corpus-based
paraphrase tool to improve EFL learners’ writing skills. Computer Assisted Language
Learning, 28(1), 22–40. https://doi.org/10.1080/09588221.2013.783873

Clark, E., August, T., Serrano, S., Haduong, N., Gururangan, S., & Smith, N. A. (2021). All
That’s `Human’ Is Not Gold: Evaluating Human Evaluation of Generated Text.
Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics
and the 11th International Joint Conference on Natural Language Processing (Volume 1:
Long Papers), 7282–7296. https://doi.org/10.18653/v1/2021.acl-long.565

Clarke, O., Chan, W. Y. D., Bukuru, S., Logan, J., & Wong, R. (2022). Assessing knowledge of
and attitudes towards plagiarism and ability to recognize plagiaristic writing among
university students in Rwanda. Higher Education. https://doi.org/10.1007/s10734-022-
00830-y

Cope, B., Kalantzis, M., & Searsmith, D. (2021). Artificial intelligence for education: Knowledge
and its assessment in AI-enabled learning ecologies. Educational Philosophy and
Theory, 53(12), 1229–1245. https://doi.org/10.1080/00131857.2020.1728732

Curtis, G. J. (2022). Trends in plagiarism and cheating prevalence: 1990-2020 and beyond. In
D. Rettinger & T. Bertram Gallant (Eds.), Cheating academic integrity (pp. 11–44).
Jossey-Bass. https://www.wiley.com/en-
au/Cheating+Academic+Integrity%3A+Lessons+from+30+Years+of+Research-p-
9781119868170

Dale, R. (2021). GPT-3: What’s it good for? Natural Language Engineering, 27(1), 113–118.
https://doi.org/10.1017/S1351324920000601

Dale, R., & Viethen, J. (2021). The automated writing assistance landscape in 2021. Natural
Language Engineering, 27(4), 511–518. https://doi.org/10.1017/S1351324921000164

Dawson, P. (2020). Cognitive Offloading and Assessment. In M. Bearman, P. Dawson, R.

Ajjawi, J. Tai, & D. Boud (Eds.), Re-imagining University Assessment in a Digital World
(pp. 37–48). Springer International Publishing. https://doi.org/10.1007/978-3-030-41956-
1_4

Dehouche, N. (2021). Plagiarism in the age of massive Generative Pre-trained Transformers

(GPT-3). Ethics in Science and Environmental Politics, 21, 17–23.
https://doi.org/10.3354/esep00195

Dinneen, C. (2021). Students’ use of digital translation and paraphrasing tools in written
assignments on Direct Entry English Programs. English Australia Journal, 37(1), 40–51.
https://doi.org/10.3316/informit.748262877947586

17

17
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

Eaton, S. E., Mindzak, M., & Morrison, R. (2021, June 29). Artificial Intelligence, Algorithmic
Writing & Educational Ethics. Canadian Society for the Study of Education, Edmontan,
AB, Canada. Werklund School of Education. https://tinyurl.com/222yctbb

Fatemi, G., & Saito, E. (2020). Unintentional plagiarism and academic integrity: The challenges
and needs of postgraduate international students in Australia. Journal of Further and
Higher Education, 44(10), 1305–1319. https://doi.org/10.1080/0309877X.2019.1683521

Fitria, T. N. (2021). Grammarly as AI-powered English Writing Assistant: Students’ Alternative

for Writing English. Metathesis: Journal of English Language, Literature, and Teaching,
5(1), Article 1. https://doi.org/10.31002/metathesis.v5i1.3519

Fröhling, L., & Zubiaga, A. (2021). Feature-based detection of automated language models:
Tackling GPT-2, GPT-3 and Grover. PeerJ Computer Science, 7, e443.
https://doi.org/10.7717/peerj-cs.443

Fu, Q.-K., Zou, D., Xie, H., & Cheng, G. (2022). A review of AWE feedback: Types, learning
outcomes, and implications. Computer Assisted Language Learning, 0(0), 1–43.
https://doi.org/10.1080/09588221.2022.2033787

Furze, L. (2022, December 6). Can an AI critique human writing? Leon Furze.
https://leonfurze.com/2022/12/06/can-an-ai-critique-human-writing/

Gayed, J. M., Carlon, M. K. J., Oriola, A. M., & Cross, J. S. (2022). Exploring an AI-based
writing Assistant’s impact on English language learners. Computers and Education:
Artificial Intelligence, 3, 100055. https://doi.org/10.1016/j.caeai.2022.100055

Gehrmann, S., Strobelt, H., & Rush, A. M. (2019). GLTR: Statistical Detection and Visualization
of Generated Text (arXiv:1906.04043). arXiv. https://doi.org/10.48550/arXiv.1906.04043

Godwin-Jones, R. (2022). Partnering with AI: Intelligent writing assistance and instructed
language learning. Language Learning & Technology, 26(2), 5–24.
https://doi.org/10125/73474

Grammarly. (2022, August 29). Innovating the Basics: Achieving Superior Precision and Recall
in Grammatical Error Correction. Grammarly Engineering Blog.
https://www.grammarly.com/blog/engineering/innovating-the-basics/

Grijalva, T. C., Kerkvliet, J., & Nowell, C. (2006). Academic honesty and online courses. College
Student Journal, 40(1), Article 1. https://tinyurl.com/3beej9bb

Gunser, V. E., Gottschling, S., Brucker, B., Richter, S., & Gerjets, P. (2021). Can Users
Distinguish Narrative Texts Written by an Artificial Intelligence Writing Tool from Purely
Human Text? International Conference on Human-Computer Interaction, 520–527.
https://doi.org/10.1007/978-3-030-78635-9_67

18

https://ro.uow.edu.au/jutlp/vol20/iss2/07 18
 Perkins: Academic Integrity considerations of AI Large Language Models in

Henderson, M., Chung, J., Awdry, R., Mundy, M., Bryant, M., Ashford, C., & Ryan, K. (2022).
Factors associated with online examination cheating. Assessment & Evaluation in
Higher Education, 0(0), 1–15. https://doi.org/10.1080/02602938.2022.2144802

Hern, A. (2022, December 4). AI bot ChatGPT stuns academics with essay-writing skills and
usability. The Guardian. https://www.theguardian.com/technology/2022/dec/04/ai-bot-
chatgpt-stuns-academics-with-essay-writing-skills-and-usability

Hutson, M. (2022). Could AI help you to write your next paper? Nature, 611(7934), 192–193.
https://doi.org/10.1038/d41586-022-03479-w

Ippolito, D., Duckworth, D., Callison-Burch, C., & Eck, D. (2020). Automatic Detection of
Generated Text is Easiest when Humans are Fooled (arXiv:1911.00650). arXiv.
https://doi.org/10.48550/arXiv.1911.00650

Jenkins, B. D., Golding, J. M., Le Grand, A. M., Levi, M. M., & Pals, A. M. (2022). When
Opportunity Knocks: College Students’ Cheating Amid the COVID-19 Pandemic.
Teaching of Psychology, 00986283211059067.
https://doi.org/10.1177/00986283211059067

Kaqinari, T., Makarova, E., Audran, J., Döring, A. K., Göbel, K., & Kern, D. (2021). The switch to
online teaching during the first COVID-19 lockdown: A comparative study at four
European universities. Journal of University Teaching & Learning Practice, 18(5).
https://doi.org/10.53761/1.18.5.10

Köbis, N., & Mossink, L. D. (2021). Artificial intelligence versus Maya Angelou: Experimental
evidence that people cannot differentiate AI-generated from human-written poetry.
Computers in Human Behavior, 114, 106553. https://doi.org/10.1016/j.chb.2020.106553

Kozma, R. B. (1991). Computer-based writing tools and the cognitive needs of novice writers.
Computers and Composition, 8(2), 31–45. https://doi.org/10.1016/8755-4615(91)80046-
G

Kumar, R., Mindzak, M., Eaton, S. E., & Morrison, R. (2022). AI & AI: Exploring the
contemporary intersections of artificial intelligence and academic integrity. Canadian
Society for the Study of Higher Education Annual Conference, Online. Werklund School
of Education. https://tinyurl.com/ycknz8fd

Kumar, R., Mindzak, M., & Racz, R. (2022). Who Wrote This? The Use of Artificial Intelligence
in the Academy. http://hdl.handle.net/10464/16532

Lancaster, T., & Cotarlan, C. (2021). Contract cheating by STEM students through a file sharing
website: A Covid-19 pandemic perspective. International Journal for Educational
Integrity, 17(1), Article 1. https://doi.org/10.1007/s40979-021-00070-0

19

19
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

Langston, J. (2021, November 2). New Azure OpenAI Service combines access to powerful
GPT-3 language models with Azure’s enterprise capabilities. The AI Blog.
https://blogs.microsoft.com/ai/new-azure-openai-service/

Lanier, M. M. (2006). Academic Integrity and Distance Learning. Journal of Criminal Justice
Education, 17(2), Article 2. https://doi.org/10.1080/10511250600866166

Lee, M., Liang, P., & Yang, Q. (2022). CoAuthor: Designing a Human-AI Collaborative Writing
Dataset for Exploring Language Model Capabilities. CHI Conference on Human Factors
in Computing Systems, 1–19. https://doi.org/10.1145/3491102.3502030

Lee, Y., Kim, T. S., Chang, M., & Kim, J. (2022). Interactive Children’s Story Rewriting Through
Parent-Children Interaction. Proceedings of the First Workshop on Intelligent and
Interactive Writing Assistants (In2Writing 2022), 62–71.
https://doi.org/10.18653/v1/2022.in2writing-1.9

Lim, K., Song, J., & Park, J. (2022). Neural automated writing evaluation for Korean L2 writing.
Natural Language Engineering, 1–23. https://doi.org/10.1017/S1351324922000298

Lin, S., Hilton, J., & Evans, O. (2022). Teaching Models to Express Their Uncertainty in Words
(arXiv:2205.14334). arXiv. https://doi.org/10.48550/arXiv.2205.14334

Luitse, D., & Denkena, W. (2021). The great Transformer: Examining the role of large language
models in the political economy of AI. Big Data & Society, 8(2), 20539517211047736.
https://doi.org/10.1177/20539517211047734

Malesky, L. A., Baley, J., & Crow, R. (2016). Academic Dishonesty: Assessing the Threat of
Cheating Companies to Online Education. College Teaching, 64(4), 178–183.
https://doi.org/10.1080/87567555.2015.1133558

Marche, S. (2022, December 6). The College Essay Is Dead. The Atlantic.
https://www.theatlantic.com/technology/archive/2022/12/chatgpt-ai-writing-college-
student-essays/672371/

McKnight, L. (2021). Electric Sheep? Humans, Robots, Artificial Intelligence, and the Future of
Writing. Changing English, 28(4), 442–455.
https://doi.org/10.1080/1358684X.2021.1941768

Miller, A., & Young-Jones, A. D. (2012). Academic integrity: Online classes compared to face-to-
face classes. Journal of Instructional Psychology, 39(3–4), Article 3–4.
https://tinyurl.com/mrxnp9zy

Nazari, N., Shabbir, M. S., & Setiawan, R. (2021). Application of Artificial Intelligence powered
digital writing assistant in higher education: Randomized controlled trial. Heliyon, 7(5),
e07014. https://doi.org/10.1016/j.heliyon.2021.e07014

20

https://ro.uow.edu.au/jutlp/vol20/iss2/07 20
 Perkins: Academic Integrity considerations of AI Large Language Models in

Nobles, S., & Paganucci, L. (2015). Do Digital Writing Tools Deliver? Student Perceptions of
Writing Quality Using Digital Tools and Online Writing Environments. Computers and
Composition, 38, 16–31. https://doi.org/10.1016/j.compcom.2015.09.001

ONeill, R., & Russell, A. (2019). Stop! Grammar time: University students’ perceptions of the
automated feedback program Grammarly. Australasian Journal of Educational
Technology, 35(1), Article 1. https://doi.org/10.14742/ajet.3795

OpenAI. (n.d.). ChatGPT. Retrieved 21 December 2022, from https://chat.openai.com

OpenAi. (n.d.). DALL·E 2. DALL·E 2. Retrieved 25 November 2022, from

https://openai.com/dall-e-2/

OpenAI. (2022, November 30). ChatGPT: Optimizing Language Models for Dialogue. OpenAI.
https://openai.com/blog/chatgpt/

Oppenlaender, J. (2022). A Taxonomy of Prompt Modifiers for Text-To-Image Generation

(arXiv:2204.13988). arXiv. https://doi.org/10.48550/arXiv.2204.13988

Palmquist, M. (2003). A brief history of computer support for writing centers and writing-across-
the-curriculum programs. Computers and Composition, 20(4), 395–413.
https://doi.org/10.1016/j.compcom.2003.08.013

Park, J.-H., & Yang, I. Y. (2020). Utilizing an AI-Based Grammar Checker in an EFL Writing
Classroom. Academic Journal of Applied Linguistics, 36(1), 97–120.
https://doi.org/10.17154/kjal.2020.3.36.1.97

Parkinson, A. L., Hatje, E., Kynn, M., Kuballa, A. V., Donkin, R., & Reinke, N. B. (2022).
Collusion is still a tricky topic: Student perspectives of academic integrity using
assessment-specific examples in a science subject. Assessment & Evaluation in Higher
Education, 47(8), 1416–1428. https://doi.org/10.1080/02602938.2022.2040947

Perkins, M., Basar Gezgin, U., & Gordon, R. (2019). Plagiarism in higher education:
Classification, causes and controls. Pan-Pacific Management Science, 2, 3–21.
https://doi.org/10.13140/RG.2.2.20694.11841

Perkins, M., Gezgin, U. B., & Roe, J. (2018). Understanding the Relationship between
Language Ability and Plagiarism in Non-Native English Speaking Business Students.
Journal of Academic Ethics, 16(4), Article 4. https://doi.org/10.1007/s10805-018-9311-8

Prentice, F. M., & Kinden, C. E. (2018). Paraphrasing tools, language translation tools and
plagiarism: An exploratory study. International Journal for Educational Integrity, 14(1),
11. https://doi.org/10.1007/s40979-018-0036-7

Price, M. (2002). Beyond ‘Gotcha!’: Situating Plagiarism in Policy and Pedagogy. College
Composition and Communication, 54(1), 88–115. https://doi.org/10.2307/1512103

21

21
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

QuillBot. (n.d.-a). Can I use QuillBot for plagiarism? QuillBot Help Center. Retrieved 16
November 2022, from https://help.quillbot.com/hc/en-us/articles/1500007485841-Can-I-
use-QuillBot-for-plagiarism-

QuillBot. (n.d.-b). Is using a paraphrasing tool plagiarizing? QuillBot Help Center. Retrieved 16
November 2022, from https://help.quillbot.com/hc/en-us/articles/4408075614103-Is-
using-a-paraphrasing-tool-plagiarizing-

Rapanta, C., Botturi, L., Goodyear, P., Guàrdia, L., & Koole, M. (2021). Balancing Technology,
Pedagogy and the New Normal: Post-pandemic Challenges for Higher Education.
Postdigital Science and Education, 3(3), 715–742. https://doi.org/10.1007/s42438-021-
00249-1

Reedy, A., Pfitzner, D., Rook, L., & Ellis, L. (2021). Responding to the COVID-19 emergency:
Student and academic staff perceptions of academic integrity in the transition to online
exams at three Australian universities. International Journal for Educational Integrity,
17(1), Article 1. https://doi.org/10.1007/s40979-021-00075-9

Risko, E. F., & Gilbert, S. J. (2016). Cognitive Offloading. Trends in Cognitive Sciences, 20(9),
676–688. https://doi.org/10.1016/j.tics.2016.07.002

Roe, J., & Perkins, M. (2022). What are Automated Paraphrasing Tools and how do we address
them? A review of a growing threat to academic integrity. International Journal for
Educational Integrity, 18(1), Article 1. https://doi.org/10.1007/s40979-022-00109-w

Rogerson, A. M., & McCarthy, G. (2017). Using Internet based paraphrasing tools: Original
work, patchwriting or facilitated plagiarism? International Journal for Educational
Integrity, 13(1), 2. https://doi.org/10.1007/s40979-016-0013-y

Sanh, V., Webson, A., Raffel, C., Bach, S. H., Sutawika, L., Alyafeai, Z., Chaffin, A., Stiegler, A.,
Scao, T. L., Raja, A., Dey, M., Bari, M. S., Xu, C., Thakker, U., Sharma, S. S.,
Szczechla, E., Kim, T., Chhablani, G., Nayak, N., … Rush, A. M. (2022). Multitask
Prompted Training Enables Zero-Shot Task Generalization (arXiv:2110.08207). arXiv.
https://doi.org/10.48550/arXiv.2110.08207

Solaiman, I., Brundage, M., Clark, J., Askell, A., Herbert-Voss, A., Wu, J., Radford, A., Krueger,
G., Kim, J. W., Kreps, S., McCain, M., Newhouse, A., Blazakis, J., McGuffie, K., &
Wang, J. (2019). Release Strategies and the Social Impacts of Language Models
(arXiv:1908.09203). arXiv. https://doi.org/10.48550/arXiv.1908.09203

Sparrow, J. (2022, November 18). ‘Full-on robot writing’: The artificial intelligence challenge
facing universities. The Guardian. https://www.theguardian.com/australia-
news/2022/nov/19/full-on-robot-writing-the-artificial-intelligence-challenge-facing-
universities

22

https://ro.uow.edu.au/jutlp/vol20/iss2/07 22
 Perkins: Academic Integrity considerations of AI Large Language Models in

St-Onge, C., Ouellet, K., Lakhal, S., Dubé, T., & Marceau, M. (2022). COVID-19 as the tipping
point for integrating e-assessment in higher education practices. British Journal of
Educational Technology, 53(2), Article 2. https://doi.org/10.1111/bjet.13169

Strobl, C., Ailhaud, E., Benetos, K., Devitt, A., Kruse, O., Proske, A., & Rapp, C. (2019). Digital
support for academic writing: A review of technologies and pedagogies. Computers &
Education, 131, 33–48. https://doi.org/10.1016/j.compedu.2018.12.005

Stuber-McEwen, D., Wiseley, P., & Hoggatt, S. (2009). Point, click, and cheat: Frequency and
type of academic dishonesty in the virtual classroom. Online Journal of Distance
Learning Administration, 12(3), Article 3. https://tinyurl.com/2awz7b7y

Sun, Y.-C. (2013). Do journal authors plagiarize? Using plagiarism detection software to
uncover matching text across disciplines. Journal of English for Academic Purposes,
12(4), 264–272. https://doi.org/10.1016/j.jeap.2013.07.002

Tertiary Education Quality and Standards Agency. (2020, October 28). What is academic
integrity? Tertiary Education Quality and Standards Agency.
https://www.teqsa.gov.au/what-academic-integrity

Vargo, D., Zhu, L., Benwell, B., & Yan, Z. (2021). Digital technology use during COVID-19
pandemic: A rapid review. Human Behavior and Emerging Technologies, 3(1), 13–24.
https://doi.org/10.1002/hbe2.242

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., &
Polosukhin, I. (2017). Attention is All you Need. Advances in Neural Information
Processing Systems, 30.
https://proceedings.neurips.cc/paper/2017/hash/3f5ee243547dee91fbd053c1c4a845aa-
Abstract.html

Wahle, J. P., Ruas, T., Foltýnek, T., Meuschke, N., & Gipp, B. (2022). Identifying Machine-
Paraphrased Plagiarism. In M. Smits (Ed.), Information for a Better World: Shaping the
Global Future (pp. 393–413). Springer International Publishing.
https://doi.org/10.1007/978-3-030-96957-8_34

Wahle, J. P., Ruas, T., Kirstein, F., & Gipp, B. (2022). How Large Language Models are
Transforming Machine-Paraphrased Plagiarism. ArXiv Preprint ArXiv:2210.03568.
https://arxiv.org/pdf/2210.03568.pdf

Wahle, J. P., Ruas, T., Meuschke, N., & Gipp, B. (2021). Are neural language models good
plagiarists? A benchmark for neural paraphrase detection. 2021 ACM/IEEE Joint
Conference on Digital Libraries (JCDL), 226–229.
https://doi.org/10.1109/JCDL52503.2021.00065

Walsh, L. L., Lichti, D. A., Zambrano-Varghese, C. M., Borgaonkar, A. D., Sodhi, J. S., Moon,
S., Wester, E. R., & Callis-Duehl, K. L. (2021). Why and how science students in the

23

23
 Journal of University Teaching & Learning Practice, Vol. 20 [2023], Iss. 2, Art. 07

United States think their peers cheat more frequently online: Perspectives during the
COVID-19 pandemic. International Journal for Educational Integrity, 17(1), Article 1.
https://doi.org/10.1007/s40979-021-00089-3

Warschauer, M., & Grimes, D. (2008). Automated Writing Assessment in the Classroom.
Pedagogies: An International Journal, 3(1), 22–36.
https://doi.org/10.1080/15544800701771580

Watson, G., & Sottile, J. (2010). Cheating in the Digital Age: Do Students Cheat More in Online
Courses? Online Journal of Distance Learning Administration, 13(1).
https://tinyurl.com/yckenh4f

Wilder, N., Weßels, D., Gröpler, J., Klein, A., & Mundorf, M. (2021). Who is Responsible for
Integrity in the Age of Artificial Intelligence? An Analysis Using the Example of Academic
Writing. European Conference on Academic Integrity and Plagiarism 2021, 179–181.
https://tinyurl.com/2s4r5dyr

Zhang, Q., Wang, D. Y., & Voelker, G. M. (2014). DSpin: Detecting Automatically Spun Content
on the Web. Proceedings 2014 Network and Distributed System Security Symposium.
Network and Distributed System Security Symposium, San Diego, CA.
https://doi.org/10.14722/ndss.2014.23004

Zhang, S., Roller, S., Goyal, N., Artetxe, M., Chen, M., Chen, S., Dewan, C., Diab, M., Li, X.,
Lin, X. V., Mihaylov, T., Ott, M., Shleifer, S., Shuster, K., Simig, D., Koura, P. S., Sridhar,
A., Wang, T., & Zettlemoyer, L. (2022). OPT: Open Pre-trained Transformer Language
Models (arXiv:2205.01068). arXiv. https://doi.org/10.48550/arXiv.2205.01068

Zhao, X. (2022). Leveraging Artificial Intelligence (AI) Technology for English Writing:
Introducing Wordtune as a Digital Writing Assistant for EFL Writers. RELC Journal.
https://doi.org/10.1177/00336882221094089

24

https://ro.uow.edu.au/jutlp/vol20/iss2/07 24