See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/374935100

Generative AI tools and assessment: Guidelines of the world's top-ranking

universities

Article in Computers and Education Open · October 2023

DOI: 10.1016/j.caeo.2023.100151

CITATIONS READS

4 349

3 authors:

Benjamin Luke Moorhouse Marie Yeo

Hong Kong Baptist University SEAMEO Regional Language Centre
85 PUBLICATIONS 2,056 CITATIONS 28 PUBLICATIONS 97 CITATIONS

SEE PROFILE SEE PROFILE

Yuwei Wan
Hong Kong Baptist University
5 PUBLICATIONS 8 CITATIONS

SEE PROFILE

All content following this page was uploaded by Marie Yeo on 02 November 2023.

The user has requested enhancement of the downloaded file.

 Computers and Education Open 5 (2023) 100151

Contents lists available at ScienceDirect

Computers and Education Open

journal homepage: www.sciencedirect.com/journal/computers-and-education-open

Generative AI tools and assessment: Guidelines of the world’s

top-ranking universities
Benjamin Luke Moorhouse a, *, Marie Alina Yeo b, Yuwei Wan a
a
Department of Education Studies, Hong Kong Baptist University, Hong Kong, China
b
Senior Language Specialist, SEAMEO RELC, Singapore

A R T I C L E I N F O A B S T R A C T

Keywords: The public release of generative artificial intelligence (GAI) tools (e.g., ChatGPT) has had a disruptive effect on
Generative artificial intelligence the assessment practices of higher education institutions (HEIs) worldwide. Concerns have largely been asso­
Assessment guidelines ciated with academic integrity, cheating and plagiarism. HEIs have had to develop guidelines in response to GAI.
Higher education
As many of these guidelines were developed in haste and could affect a large number of instructors and students,
ChatGPT
Academic integrity
there is a need to examine their content, coverage and suitability. This review examines the extent to which the
world’s 50 top-ranking HEIs have developed or modified their assessment guidelines to address GAI use and,
where guidelines exist, the primary content and advice given to guide instructors in their GAI assessment design
and practices. The findings show that just under half of the institutions have developed publicly available
guidelines. The guidelines cover three main areas: academic integrity, advice on assessment design and
communicating with students. Amongst the suggestions for teachers on assessment design, two appear partic­
ularly pertinent in helping develop effective assessment tasks and developing learners’ AI literacy: first, running
assessment tasks through GAI to check the extent to which the tool can accomplish the task and, second, having
students use GAI as part of the assessment process. Overall, the review suggests that HEIs have come to accept the
use of GAI and drafted assessment guidelines to advise instructors on its use. In the article, we argue that it may
be beneficial to embrace GAI as a part of the assessment process since this is the reality of today’s educational
and job landscape. This will require instructors to develop a new competence - generative artificial intelligence
assessment literacy - which is conceptualised in this article.

1. Introduction images, text, simulations, 3D objects and videos. These tools can create
unexpected outputs in response to varied and complex prompts (e.g.,
The growth of Artificial Intelligence (AI), particularly generative AI languages, instructions, questions) [6].
(GAI) such as ChatGPT and Bard, has intensified calls for guidance of its A key concern of HEIs is that students may use GAI to cheat or
use in higher education institutions (HEIs) worldwide [1,2]. This is in plagiarise their written assignments and exams [7]. This could under­
large part due to the ability of these tools to generate human-like texts mine academic integrity and ultimately the reputation of HEIs. In
that are difficult to detect even by experts [3], and the incredibly fast addition, scholars have argued that students could become over-reliant
adoption of these technologies by students. Indeed, The Times Newspaper on GAI, leading to a decline in their writing and critical thinking skills
reported that nearly half of the students at Cambridge University in the [8] and negatively affect the quality of education and student learning
United Kingdom admitted to using ChatGPT in their studies [4] while outcomes [9].
Forbes magazine claimed that one in five, or 20 % of college students in However, other scholars have argued that GAI can bring benefits to
the United States, admitted the same [5]. These figures are likely to education (e.g., [10,11]). For example, Hwang and Chen [12] provide
increase as GAI functionality becomes embedded in word processors and six possible roles for GAI in education: teacher/tutor; student/tutee;
presentation software (e.g., Microsoft Co-Pilot). GAI represents a dra­ learning peer/ partner; domain expert; administrator; and learning tool.
matic advancement from previous AI models. GAI leverages deep Indeed, traditional (non-generative) AI tools have been found to have a
learning models to generate human-like content, including audio, code, number of positive utilities in higher education. Crompton and Burke

* Corresponding author.
E-mail address: [email protected] (B.L. Moorhouse).

https://doi.org/10.1016/j.caeo.2023.100151
Received 7 July 2023; Received in revised form 14 September 2023; Accepted 1 October 2023
Available online 23 October 2023
2666-5573/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

[13] in their systematic review of AI use in HEIs from 2016 to 2022 integrity as “the expectation that teachers, students, researchers and all
found various pedagogical benefits of using AI. These included members of the academic community act with: honesty, trust, fairness,
providing more personalised feedback on assignments, helping in­ respect and responsibility”. Academic dishonesty, in contrast, refers to
structors generate questions and tests (e.g., [14,15]), accurately pre­ "a range of behaviours involving intentional violation of academic rules
dicting academic performance (e.g. [16]), and providing tailored and for personal gain, such as plagiarism, lying, and falsifications’’ ([26], p.
personal academic assistance (e.g. [17]). Similarly, Celik et al. [18] 100,054). TEQSA [25] further lists plagiarism, recycling or resubmitting
found that AI can help teachers define and support students’ needs, work, fabricating information, collusion, exam cheating, contract
provide immediate student feedback, and conduct automatic essay cheating and impersonation as examples of academic dishonesty.
scoring. Within the literature, plagiarism appears to be the most frequent
Despite the proven and potential benefits of AI in education, the form of academic dishonesty [27]. Yet plagiarism itself is a multifaceted
disruptive nature of the general release of GAI technologies to common concept with diverse interpretations and definitions. One form of
assessment practices has led to widespread calls for HEIs to develop plagiarism is “textual plagiarism”, which refers to “the use of words
comprehensive guidelines pertaining to their use (e.g. [19,20]). For and/or ideas from another source, without proper attribution,” and this
example, Rudolph et al. [19] suggests the HEIs, "develop policies and may be intentional “prototypical plagiarism” or unintentional “patch­
clear, easy-to-understand guidelines for the use of language models in writing” ([28], p. 276). Another act of plagiarism is to list sources in the
learning and teaching - the guidelines should include information on the reference list that the student has not actually read [26]. Based on these
proper use of these tools and the consequence of cheating" (p.356). conceptualizations of plagiarism, deliberately using GAI to construct
Early reports on the reactions of HEIs to GAI suggested a range of assessment responses that include citations and references would still
responses to their use. Some universities banned the use of GAI tools in constitute textual plagiarism [11]. In addition to plagiarism, students’
their academic programmes, with the use of AI seen as cheating [7,21], use of GAI to complete assignments might also be deemed as collusion,
while others have guidelines allowing their use, as long as that use is contract cheating, and impersonation. Although it is common for aca­
declared and acknowledged [22]. Clearly, the development of GAI demic integrity policies to include clauses pertaining to these aspects,
technologies has complicated the HEI assessment landscape and blurred often they refer to collusion with “other students” or “commissioning to
the line between acceptable and unacceptable practices. The lack of someone else” (e.g., [25]). These clauses were developed before the
clarity could be troubling for instructors and learners, who look to emergence of GAI and may require HEIs to reflect this change in any
institutional guidelines to guide their practices [11,23]. state, “A culture subsequent guidelines. GAI likely opens up the possibility of contract
of academic integrity in institutions is enhanced if policies and pro­ cheating to vastly more students [29].
cedures are carefully written with all key stakeholders in mind” (p. 6). To promote academic integrity and deter academic dishonesty, HEIs
Although HEIs were initially “caught off guard” by the development have generally taken a two-pronged approach: detection and preven­
of GAI [20], an increasing number of them have begun to develop tion. Plagiarism detection software such as iThenticate and Turnitin are
guidelines pertaining to their use in teaching, learning and assessment. commonly used to “check” students’ assessment submissions. How well
We believe there can be much to learn from different HEIs’ responses to these tools can detect original texts generated by AI remains question­
GAI, allowing for a collective shaping of clear and consistent guidelines able [30]. Some tools such as GPTZero, Turnitin, ZeroGPT and Winston.
regarding GAI, academic integrity and assessments. AI claim to be able to detect text generated by GAI. Despite these claims,
Within this rapidly advancing context, this review article aims, many scholars have questioned the accuracy of these tools (e.g. [31]).
firstly, to identify the extent to which the world’s top ranking HEIs have Kohnke et al. [11] note that the use of these tools may “lead to a game of
developed or modified guidelines specifically to address the complex­ cat and mouse” (p.544) with GAI developers and detection tools
ities posed by the use of GAI tools in assessment tasks. Secondly, where constantly coming up with ways to out-pace each other, leaving users
guidelines exist, we seek to identify the primary content and advice with no assurance of reliable detection.
contained within them that can help guide instructors regarding GAI and
assessment. In this article, we define guidelines as non-mandatory rec­ 1.2. Assessment design and academic integrity in the GAI world
ommendations, interpretations, administrative instructions, best prac­
tice guidance, or frameworks in which to operate. Policies, on the other Assessment design and processes have been another area of debate
hand, are mandatory, prescriptive and have far-reaching aims (e.g. since the development of GAI [30]. These debates have focused on the
institutional reputation, legal compliance, operational efficiency) [24], integrity of the assessment process and the qualities of effective assess­
so they generally take a long time to develop and be approved. Based on ment may be compromised if students use GAIs to complete their
this, it is likely that most of the examples from the websites we surveyed assessment tasks. For example, Yeo [32] suggested that the validity,
may, in effect, be guidelines, even if they are referred to as policies. reliability, and fairness of the assessment process may be compromised
We explicitly focus on GAI guidelines, as they offer tangible exam­ if students are assessed based on work written partly or wholly by GAI
ples of advice and suggestions that can be adopted by instructors in HEIs. tools. As an illustration, if instructors assign a written essay to test the
At the same time, we hope that the findings of this paper offer practical writing sub-skills of topic development, organisation and coherence,
suggestions for designing assessment in higher education in the GAI lexical resource and grammatical range and accuracy, it is with the
world. The review serves as a reference for HEIs in the process of expectation that students will construct and produce a response on their
designing or redesigning their guidelines around academic integrity and own so that the instructors can assess them against the assessment
assessments. criteria. However, if students use a GAI tool to complete the task, they
will not be demonstrating these skills but may instead be using other
1.1. Academic integrity and dishonesty in the GAI world skills, such as the ability to write initial and further prompts to guide the
tool to produce and refine the generated output. In addition, some stu­
As previously stated, the primary concerns regarding AI and GAI and dents may have access to specialist GAI tools, or more advanced tools (e.
assessments in HEI have been the issue of academic integrity and aca­ g. Chatgpt 3.5 vs GPT-4), and others do not. This can affect the reliability
demic dishonesty [2,7,19]. Various institutions and government and fairness of an assessment as students do not have equal access to
agencies have policies regarding academic integrity and provide guid­ resources that can help them to succeed in the assessment [33].
ance on the implications of academic dishonesty. In order to oper­ An initial response to the above issues and risks by many HEIs was to
ationalize these policies, clear definitions are needed. For example, The ban the use of GAI [34] or control the environment in which students did
Australian government’s Tertiary Education Quality and Standards their assignments and to return to proctored pen-and-paper tests [21].
Agency (TEQSA) [25] provides the following definition of academic Such measures were supported by evidence that the environment where

2
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

students do assignments, at home or in a controlled classroom envi­ 2.1. Locating GAI guidelines
ronment, has an impact on academic integrity. For example, Yuzici et al.
[35] noted, “The more the students work out-of-class or perform To locate the GAI guideline documents in each university’s website,
take-home exams and assessment, the greater the risk to their academic the researchers visited all the top 50 universities’ official websites and
honesty” (p. 222). Janke’s [26] study found that students were more conducted manual searches. The initial keywords used for the search
likely to commit acts of academic dishonesty for assessments not done included “AI policy”, “Generative AI policy”, “ChatGPT policy”, “AI
under proctored, timed conditions. guidelines”, “Generative AI guidelines’’, “ChatGPT guidelines”, “AI
However, due to the known limitations of in-class assessments (e.g. guide”, “Generative AI guide”, and “ChatGPT guide”. Additionally, some
limited scope, test-taking anxiety, lack of differentiation, decontextu­ related terms, such as “academic integrity”, “academic honesty” and
alization, irrelevance to today’s workplace) [36], there has been a move “plagiarism”, were identified and incorporated into the search to ensure
away from in-class assessments towards take-home assessments in comprehensive coverage. After conducting the initial keyword search,
recent years [28]. Although this has opened up the kinds of assignment we also took additional steps to ensure that no relevant information was
tasks instructors can assign, with various recommendations proposed to missed. We visited specific sections or columns on each university’s
increase the authenticity of assignments and diversity of assignment official website that could potentially contain GAI guidelines (e.g. the
tasks (e.g. portfolios, case studies) [37,38], there is still a dominance of websites of teaching and learning centres). Of the 50 universities, 30
written essays in many HEI programmes and courses. Common were found to have guidelines related to generative AI on their official
take-home assessments include responding to questions, conducting a websites (60 %). The websites were extracted for analysis. The search
literature review, writing an argumentative or analytical essay, pro­ and extraction were conducted on 15th June 2023. Only publicly
ducing presentation slides, or writing research reports [39]. These kinds available documents or websites were included due to access
of writing tasks can be accomplished by GAI tools [3]. Indeed, Cotton restrictions.
et al. [29] demonstrated the ability of ChatGPT to construct an academic
text by publishing the results as an article in the journal, Innovations in
2.2. Applying inclusion criteria
Education and Teaching International with a discussion on the implica­
tions of the tool for academic integrity. Similarly, King [40] published a
To ensure the websites were relevant to our review aims, we applied
conversation he had with ChatGPT as an editorial in Cellular and Mo­
the following inclusion criteria:
lecular Bioengineering to highlight the capabilities of the tool.
Scholars have suggested then that we need to reconsider the kinds of
• The audience must be instructors (exclude guidelines for researchers
assessment tasks instructors assign to make them ‘AI-resistant’ by
and academic publishing)
reducing the likelihood that GAI can complete the whole assignment
• The genre must be guidelines (exclude blogs, notes, memos and
task. For example, Rudolph et al. [19] suggest that students can be
news)
tasked with analysing videos or images, analysising in-class discussions,
• The guidelines must be issued at the university level (exclude faculty,
analysising long texts that do not fit in a prompt, writing about recent
school, and department level)
events that are not in the training data of GAI tools and writing about
topics that are highly specific, niche or personal to the students. While
After applying the inclusion criteria, seven universities were
there is an increasing amount of advice available to instructors on how
excluded. The remaining 23 university guidelines are included in the
to modify their assignment tasks in the GAI world (e.g. blogs, newslet­
review (See Appendix for the list of universities and corresponding
ters), many instructors will look to their institutions for guidance and
websites included in the review. Please note that the websites are active,
direction regarding GAI [11]. Therefore, it is important to understand
and content may have changed since they were extracted for this study).
the kinds of suggestions HEIs are providing to their instructors.
The study consists of policies or guidelines from 15 universities located
in the USA; 4 in the UK; 2 in Canada; and 1 in Japan and Australia
2. Methodology
respectively.

This review seeks to investigate the extent to which HEIs have

developed guidelines in response to the emergence of GAI tools and 2.3. Conducting the review
identify the components of such policies. The following research ques­
tions guided the study: Due to the rapid nature of the development of the guidelines, it was
important to conduct two levels of analysis: context level and content
1. To what extent have the world’s top-ranking HEIs developed or level.
modified assessment guidelines specifically to address the use of GAI
tools in assessment tasks? 2.3.1. Context-level
2. Where guidelines exist, what is the primary content of the guidelines At the context level, data pertaining to the issuer of the guidelines
and the advice contained within the guidelines that can help guide (Table 1); issue or version date (Table 2); and tools (Table 3) are
instructors regarding GAI and assessment? mentioned. It can be seen from Table 1 that the primary issuer of
guidelines pertaining to the use of GAI in assessments is university
To ensure we selected guidelines in a systematic and efficient way, centres for teaching and learning (Note that University of Toronto
we decided to use the Times Higher Education (THE) World University (UOFT), University of California, Berkeley (UCB) and. University of
Rankings 2023 (https://www.timeshighereducation.com/world-univer
sity-rankings) to identify universities for inclusion. We decided to Table 1
include the top 50 ranked universities in the review. While this index is Issuer of guidelines pertaining to Assessment and AI.
well regarded, we recognise using the THE does present limitations. issuer quantity
First, the rankings tend to favour English-speaking countries in the
Centre for teaching, learning or innovation 15
Global North. Second, the guidelines may not be representative of all Office of Provost 3
universities globally. However, despite these limitations the selection office of ethics/academic integrity/community standard 2
criteria allowed us to conduct the review rapidly and systematically. Centre for technology 2
Below the process of conducting the review is presented. Senate 1
Not stated 3

3
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

Table 2 learning), we did not see any clear differences. The issuer may therefore
Issue month of guidelines pertaining to Assessment and AI. reflect the university’s decision-making and communication process and
Issue month Quantity (HEI) structure rather than a distinction between policies and guidelines. As
Table 2 shows, most guidelines do not have a specific date of issue with
Without specific date 15
January 2023 1 (PU) one or two issued each month between January 2023 and June 2023.
February 2023 2 (SU, UCL) Table 3 shows that more than half of the guidelines (57 %) only mention
March 2023 2 (UCB, ICL) ChatGPT and not other types of GAI tools. Many of the others mention
April 2023 1 (UTokyo) other tools, but use ChatGPT as the example tool throughout the
May 2023 1 (UOFT)
June 2023 1 (Edin.)
guidelines.

2.3.2. Content-level
At the content level, the process was more complex. First, two of the
Table 3
researchers read five websites independently and coded the content
GAI tools mentioned in the guidelines pertaining to Assessment and AI.
inductively. They recorded their codes and used them to create a
Tools mentioned Quantity framework for analysis. The two researchers met and discussed their
Only ChatGPT 13 initial coding and framework. They found consistency in their process
Not only ChatGPT (also include other tools, e.g. DALL-E, CoPilot, etc.) 10 and coding and discussed some minor discrepancies. They created a
single framework comprising of nine fields: authorship; acknowledge­
ment; plagiarism; advice on assessment design and tasks; detection of
British Columbia (UBC) had two issuers of guidelines). Interestingly,
when we compared the scope and content of the documents issued by GAI use; responsible agent; proper use; improper use; and communica­
tion with students. Themes within fields and guiding questions were also
different issuers (e.g., senate, provost, centres for teaching and

Fig. 1. Main Fields in GAI Guidelines.

4
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

included to aid the review. The two researchers then independently several GAI paraphrasing tools (e.g., Quilbot). UCLA reminded its in­
sorted, coded and extracted both quantitative (if the field was structors to be mindful of such behaviour. In addition, the guidelines of
mentioned in the website) and qualitative information (what the University College London (UCL) and Duke University (DU) suggested
guidelines included pertaining to the field) from the remaining websites. that students may use AI-generated content as a source of information or
Additional fields or themes were included if content did not match the inspiration but if they failed to properly cite or acknowledge the source,
nine fields above. After the researchers completed the process, they met this behaviour would be considered as plagiarism.
to discuss the outcomes. Each website was discussed one by one, with To help students better understand the importance of academic
any discrepancies discussed until agreement was reached. integrity and the consequences of violating the university’s plagiarism
At this stage, the quantitative data was collocated and subjected to policy, UCLA and University of Illinois at Urbana-Champaign (UIUC)
standard descriptive statistical analysis (quantities and percentages). suggested instructors incorporate relevant information into their course
The qualitative data was read and re-read again with data extracts syllabus. Some guidelines provided templates or examples of the policy
illustrative of the themes within the fields identified for inclusion in this language. For example, DU provided the following sample; “All work
paper. In summary, our review of the GAI guidelines on the websites of submitted in this course must be your own. Contributions from anyone
23 universities listed amongst the 50 top universities in the THE Index or anything else—including AI sources—must be properly quoted and
(2023) yielded three main areas of coverage, which can further be cat­ cited every time they are used”.
egorised as shown in Fig. 1. The findings regarding each area are pre­ When students are suspected of plagiarism, three universities
sented in detail in the next section. mentioned how instructors should respond to such behaviour. Both
University of Pennsylvania (UPenn) and Northwestern University (NU)
3. Findings suggested that instructors should refer to university’s standard of aca­
demic integrity, but Upenn also emphasised that instructors have the
3.1. Academic integrity discretion to decide whether a violation occurred or not. UCL recom­
mended that if instructors have concerns about a student’s work
Given the nature of the guidelines, they all included advice on aca­ potentially containing AI-generated content without proper documen­
demic integrity and GAI in assessments. However, not all the guidelines tation, students should be invited to “an investigatory viva to probe the
provided the same advice, nor did they include the same information. authorship” of their work.
Three main fields were identified: plagiarism, acknowledgement of GAI
and detection of GAI use. Below, data pertaining to each field are pre­ 3.1.2. Acknowledgement of GAI
sented (See Table 4 for the themes related to academic integrity and HEI More than half (57 %) of the universities provided guidelines on
guidelines that exhibit the themes). proper acknowledgement of GAI in students’ assignments. The two main
ways suggested are (1) acknowledging the use of GAI and (2) citing AI-
3.1.1. Plagiarism generated content. Eight universities mentioned acknowledging GAI use
Sixty percent of the university guidelines addressed plagiarism or in their policies or guidelines. Four of them only emphasised the
academic misconduct in their GAI guidelines. The analysis of the doc­ importance of disclosing the use of GAI assistance without providing
uments identified three main themes: (1) Forms of plagiarism using GAI; further guidance on how to properly document and attribute the AI-
(2) Ways to address plagiarism; and (3) Instructors’ responses to generated content. The remaining four universities offered more
plagiarism. comprehensive guidelines for acknowledging GAI use, including
Eight university guidelines identified three forms of plagiarism using mentioning the specific AI tool used and the date it was accessed (Uni­
GAI. These were copying and pasting AI-generated response, running versity of Edinburgh [Edin.], UOFT), documenting the process of using
material through multiple AI generators to avoid detection, and inade­ GAI tools (UCL, Monash University [MU]), and providing an appendix
quate documentation of the use of GAI in assessment tasks. UCB (UOFT). UOFT stated that the appendix for each assignment should
explicitly stated that “Lifting full sentences and paragraphs wholecloth, include “what tools were used, how they were used, and how the results
whether it’s from an encyclopaedia, written article, or AI-generated text from the GAI were incorporated into the submitted work”. Similarly,
creation tool, is considered plagiarism.” Similarly, Imperial College UCL pointed out that students must describe the prompts used, the
London (ICL) suggested that submitting assessments generated by AI as output, and how the output generated by GAI was changed by students.
if it was students’ own is plagiarism. University of California, Los Six universities indicated how to cite or give credit to content
Angeles (UCLA) guidelines suggested that to avoid being caught by generated by GAI following source citation academic conventions
plagiarism detection tools, some students may run their work through including in-text citation and including the GAI tool in the reference list.
For the reference style, half of them recommended students use standard
citation formats (e.g., MLA, APA, Chicago), as these organisations have
Table 4 already provided information on how to cite GAI. Given the interactive
Themes related to academic integrity and HEIs HEI guidelines that exhibit the nature of GAI tools, Edin. suggested citing AI-generated content as
themes.
“personal communication”, typically using an in-text citation only.
Themes Sub-themes Institutions UIUC also recommended instructors consider different ways of citations
Plagiarism Forms of plagiarism using UCB; ICL; UCLA for different uses, such as “ideation, inspiration, or brainstorming”.
GAI UCL; Duke; NU; Edin.; There seems to be no standard way within HE for students to currently
UToyko acknowledge GAI use in assessments.
Ways to address plagiarism Caltech; UCLA; Duke; UIUC
Instructors’ responses to UPenn; NU; UCL
plagiarism 3.1.3. Detection of GAI use
Acknowledgement of Acknowledge the use of SU; PU; UCB; UPenn; 61 % of the universities mentioned the existence of GAI detection
GAI GAI UOFT; UCL; Edin.; MU tools (e.g. GPTZero, Turnitin AI detection feature). The majority of the
Cite AI-generated content DU; CMU; Edin.; UBC; guidelines discouraged instructors from relying on the tools as a way to
UIUC; UOFT
Detection of GAI use Discourage the use of GAI ICL; UOFT; UCB; YU;
check if students have followed academic integrity policies. Reasons
detection tools Cornell; DU stated for discouraging their use included: inaccuracy in identifying AI
Edin.; UTokyo; MU generated content and privacy concerns. For example, ICL and DU
Try GAI detection tools UPenn; NU; UBC; CMU; reminded their colleagues that because AI detection tools are still in the
UIUC
infant stage, they are “unproven” and “by no means foolproof” in

5
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

detecting AI. In addition, MU suggested detection tools cannot “keep • Paste an entire assignment brief into ChatGPT and see what it
pace with the latest developments”. Some universities suggested there produces.
could be privacy and security issues if instructors input students’ work • After reviewing the result, add additional instructions into ChatGPT
into the external sites (explicitly referencing AI detection tools). UCB to try to finetune the results.
suggested this may lead to ethics, privacy, and copyright violations. The • Sometimes, poor results are due to entering insufficient prompts into
guideline of DU cautioned that the use of such tools can “signal that ChatGPT rather than because ChatGPT cannot produce a satisfactory
students should not be trusted”, therefore, breaking the trust between response to an assignment. Test this by adding further instructions
staff and students. derived from the assessment rubric, marking guide, and taught ma­
Five university guidelines suggested instructors can try GAI detec­ terial. Try a variety of combinations of instructions and see if this
tion tools. However, these guidelines also point out the importance of improves the results.
being aware of the detection tools’ limitations. All the guidelines
reminded instructors that these tools should not be the "sole factor" to As the instructions above show, some documents suggested in­
determine whether students have breached academic integrity (e.g. structors try out variations of the task instructions and prompts and
UBC). examine the results. By becoming familiar with the way the tools handle
the assessment tasks, instructors can consider how they might re-design
3.2. Advice on assessment design tasks that focus on current limitations of the tools. However, Carnegie
Mellon University (CMU) guidelines emphasised an important point,
A majority, seventeen (74 %), of the reviewed universities provided that instructors must be cautious about only focusing on current limi­
instructors with advice pertaining to the design of assessment tasks in tations, as “ChatGPT will adapt as it is used. Consequently, designing
response to the development of GAI tools. There was a recognition in the assignments around any current limitations may be a temporary solu­
documents that GAI tools required instructors to examine their current tion, but not a sustainable one”. Additionally, UOFT and University of
assessment tasks and practices. The advice provided was categorised Texas at Austin (UTA) reminded instructors that there are privacy risks if
into five themes: (1) Test assessments on GAI tools; (2) Re-design examination papers or other assessment tasks are inputted into GAI
assessment tasks; (3) Focus on process and staged assessment design; (4) tools. UOFT guidelines stated, “since exam questions are highly confi­
Incorporate AI tools in the assessment design and; (5) Use in-class as­ dential documents, you should not, in principle, input them directly into
sessments (See Table 5 for themes related to assessment design HEI GAI tools”. UTA reminded their teaching staff to review the privacy and
guidelines that exhibit the themes). data collection policies of the tools before using them to test assessment
tasks.
3.2.1. Test assessments using GAI tools
Seven universities suggested instructors test their current and pro­ 3.2.2. Re-design assessment tasks
posed assessment tasks using generative AI tools. When providing this All seventeen universities suggested different ways instructors can
advice, most guidelines specifically mentioned ChatGPT only. The re-design assessment tasks in light of the development of GAI tools. The
testing of the assessments on GAI tools allows instructors to “assess their primary advice focused on (1) designing assessments that require crea­
capabilities and limitations” (UBC) and “Familiarising yourself with the tivity and critical thinking, (2) incorporating contextual elements, (3)
AI software available in your discipline” (ICL). It also raises instructors’ designing authentic assessments, and (4) providing alternative ways for
awareness of the types of texts generated by AI so they can recognise it students to represent their knowledge beyond text.
(UPenn). MU provided its instructors with step-by-step instructions on The guidelines suggested that instructors design assessments that
how to do this: require creativity and critical thinking as they assume that GAI tools
currently struggle with and they cannot “easily replicate” (ICL) these
kinds of tasks. UOFT suggested that their colleagues design tasks that:
Ask real questions that stem from current debates in your discipline,
and let students know that you expect engaged critical thinking that is
Table 5
appropriate for the level of your students and your discipline. Encourage
Themes related to assessment design and HEI guidelines that exhibit the themes.
speculation based on evidence and reasoning, not just compilation of
Theme Sub-theme Institution existing information or expression of unsupported personal opinion.
Test Assessments ICL; UPenn; CMU; UTokyo; UTA provided a long list of tasks that they suggest AI tools could
Using Generative UBC; MU; UIUC struggle with. These include,
AI tools
Give a hug: empathy, collaboration, communication, and leadership
Re-design assessment Design assessments that PU; ICL; UOFT; CMU; MU;
tasks require creativity and UTA skills; Solve a mystery: generating questions and problem finding; and
critical thinking Tell a story: finding what’s relevant in a sea of data or applying values,
Incorporate contextual UCB; UPenn; UW; CMU; ethics, morals, or aesthetic principles to a situation.
elements UBC; MU; UIUC They advised colleagues to consider these when re-designing tasks.
Design authentic YU; CU; Cornell; MU
assessments
Seven of the universities advised colleagues to incorporate contex­
Provide alternative ways for YU; UPenn; UOFT; NU; tual elements into their assignment design. This could include asking
students to represent their CMU; UTokyo; UBC; PU; “students to connect course content, class conversations, and lived
knowledge beyond text MU; UIUC experience” (University of Washington [UW]) and "design essay and
Focus on process and Caltech; PU; CU; UPenn;
exam prompts that require close discussion or analysis of the materials
staged assessment UOFT; Cornell; NU; UW;
design CMU; MU; UIUC; UTA; used for your class, including images, video, and other media.” (UCB).
UTokyo UIUC suggested instructors “make assignments more personal, reflec­
Incorporate GAI Caltech; UCB; YU; ICL; CU; tive, specific, local, based on scenarios/facts/topics covered in class, or
tools in the UW; UBC; MU; UIUC; UTA that address more complex cognitive skills”. The guidelines emphasised
assessment process
In-class assessments Recommend the use of in- UTokyo; UBC; UIUC; UTA
that GAI tools can struggle with these kinds of tasks, while students can
class assessments be more motivated to complete tasks that they find more relevant to
Caution against the over- CMU; MU their lived experiences (UIUC, CMU). Similarly, authentic assessments,
reliance of in-class which are designed for students to apply course concepts in real world
assessments
situations or problems (Columbia University [CU]) are promoted by four

6
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

of the universities. MU suggested various kinds of authentic assessments they might incorporate GAI tools into their assessment designs. Most of
including, “case studies, exhibitions, reflective portfolios, and problem- these suggest that instructors design an assessment task that requires
based inquiries”. The MU guidelines suggested that research supports students to generate responses on tools such as ChatGPT and then
the “role of authentic assessment (when done well) in improving critique their responses. For example, the guidelines of YU recom­
engagement and preventing students from engaging in academic mended, “Engaging with ChatGPT as a tool that exists in the world and
misconduct.” They cited Sotiradou et al. [38] to support their advice. having students critically engage with what it is able to produce”. ICL
Ten universities suggested that instructors consider providing alter­ suggested students can be shown how to use ChatGPT as a formative
natives and a variety of ways for students to represent their knowledge assessment tool. CU and UW suggested that by designing assessments
beyond text. These included drawing images, making slides, facilitating that require students to use generative AI tools, they can foster digital
a discussion (Yale University [YU]), submitting slides or presentations, literacy skills. Importantly, MU reminded instructors that students
video or audio recordings, or infographics (UPenn); producing portfo­ should acknowledge their use of GAI tools. UBC reminded instructors
lios, logbooks or assessment notebooks (UOFT) and using live interviews that there may be ethical and privacy concerns if they require students to
(MU). Some universities encouraged instructors to give choice to stu­ use tools in assessment tasks.
dents on how they present evidence of their learning (e.g., YU). While
the guidelines suggested instructors consider these kinds of assessments, 3.2.5. Use in-class assessments
UBC asked colleagues to be aware that GAI tools are developing rapidly Four universities recommended the use of in-class assessments as one
and may be able to create videos and other modes of texts in the near approach to mitigate the use of GAI tools in assignments. For example,
future. UIUC advised that, “Do writing assignments in class. It may be useful to
have students compose a handwritten essay or write a shorter reflection
3.2.3. Focus on process and staged assessment design paper on a current topic in class”. Similarly, UBC suggested that,
Nearly all the documents that provided advice on assessment design, One approach to mitigating the use of ChatGPT in assignment and
13 of 17, mentioned the need for instructors to put more emphasis in assessment design is to incorporate more in-class or otherwise syn­
their assessment design on the process of completing an assessment task. chronous assignments, either written or oral, or change your current
This makes students think more about the process of completing a task. grade weighting to emphasise these.
A number of ways were suggested for how they might do this: However, they also acknowledged this strategy may not be feasible
for all classes.
• Instructors can ask students to “Submit notes they took on sources to On the opposite side, there were words of caution from both MU and
prepare their papers or presentations” (UPenn) CMU that instructors should avoid an over reliance on in-class assess­
• Instructors can add “elements such as proposals, drafts, annotation, ments, such as oral exams and presentations. MU mentioned that these
or feedback into your assignments.” (NU) assessments limit the kinds of knowledge students can demonstrate,
• Instructors can “Use more iterative processes of assessment such as while CMU reminded their colleagues that these kinds of assessments
student peer review which leads to revisions of the work.” (MU) can disadvantage certain learners. Their guidelines stated that, "one or
• Instructors can ask students to provide “a list of specific steps they more of these approaches may appear to be a simple solution, these
took, what they could have done differently, and why” (CMU) changes could raise more difficulties than they solve, particularly for
• Instructors can scaffold assignment tasks by breaking “big assign­ reasons of equity and inclusion. For example, some of these approaches
ments into smaller pieces” (UIUC) may inadvertently disadvantage non-native English speakers or students
requiring accommodations for disabilities".
Breaking larger assignments into smaller pieces was a particularly Interestingly, the majority of guidelines did not provide any guid­
popular suggestion within the guidelines. There seemed to be a few key ance on in-class assessments.
reasons for this suggestion. Notably, Cornell University (Cornell) sug­
gested that a staged assignment can lower grade anxiety, and therefore 3.3. Communication with students
reduce the chance students will consider cheating. Similarly, UTA
suggests, Nearly all the guidelines (87 %) provided advice on how instructors
Carefully scaffold assignments with time and space for students to can communicate with students about the use of GAI in assessments.
complete each step along the way, and consider whether the number of They provided advice on: 1) the channels of communication, and 2) the
time-intensive tasks might require more bandwidth than students have content of the communication (See Table 6 for themes related to
to spend. Students are more likely to utilize a tool like ChatGPT when
they are short on time. Table 6
CU emphasised the value of staged assignments in providing infor­ Themes related to communicating with students and HEI guidelines that exhibit
mation to instructors on student performance, and the opportunity it the themes.
gives for students to receive formative feedback and peer feedback. In Themes Sub-themes Institutions
addition to a focus on process and staged assignments, three universities
The channels of Include a statement in syllabi or SU; PU; YU; UPenn;
suggested instructors consider more balanced and flexible submission communication course outlines UOFT; DU; NU; MU
policies. For example, Cornell suggested instructors can offer a set Engage students in open SU; PU; UPenn;
number of late submissions, allow more time for major assignments, and discussions around the use of GAI Cornell; DU; CMU;
reduce the weight of major assignments. This, they suggested, can Edin.; UTA
Collaborate with librarians UCB; UIUC
reduce anxiety, and reduce the likelihood students will use GAI tools
2) Content of Set clear expectations CU; Cornell; UTokyo;
improperly. CMU suggested that “Students may turn to ChatGPT if they communication UBC
are feeling stressed, overwhelmed, unsupported, or out of time” and Partner with students to develop CU
therefore instructors should help provide a more balanced workload. class policies
This included timing assessments to take place outside of exam periods, Discuss the ethics and limitations UCLA; DU; UBC
of GAI
providing longer deadlines, and building in time in-class for students to Mention the importance of DU
work on assignments. originality
Highlight the importance of UTokyo
3.2.4. Incorporate GAI tools in the assessment process college learning, intellectual
struggle, and process
Ten of the universities provided suggestions to instructors on how

7
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

communicating with students and HEI guidelines that exhibit the should "Facilitate discussions with your students on the impacts of
themes). spreading disinformation or biased information, lack of regulation of
companies that develop these technologies, and other dangers".
3.3.1. Channels of communication Finally, some of the guidelines suggested that instructors focus on the
Three primary suggestions are made for how teachers should value of college learning, the intellectual struggle involved in learning
communicate with students about the use of GAI in assessments. The and the importance of process in learning. For example, the University of
first is to include a statement in syllabi or course outlines. The second is Tokyo (UTokyo) stated, “‘Emphasise the importance of the process of
to engage students in open discussions around the use of AI and the third finding an answer rather than the final answer itself”. DU suggested,
is to collaborate with librarians. Suggesting instructors inform students “Instructors can emphasise why original writing (or coding or creativity)
of their policies and expectations around the use of GAI in assessments matters and what it means to develop your own voice and ideas. Un­
seems common advice in the reviewed guidelines. Some guidelines derstanding how these skills will help them in careers and further study
stated that a statement should be included but do not provide specific in your discipline can motivate students to avoid unwanted use of AI.”
examples (e.g. UOFT, MU), however, most guidelines provided exam­ Given the difficulty of detecting GAI content and the versatility in its
ples of the statements instructors can use. For example, PU provided two abilities, it seems necessary that instructors help students realise the
statements depending on the GAI policy an instructor adopts: implications of their use of such tools in the short, medium and long
term for themselves and the wider society.
1. Intellectual honesty is vital to an academic community and for my
fair evaluation of your work. All work submitted in this course must 4. Discussion
be your own, completed in accordance with the University’s aca­
demic regulations. (https://rrr.princeton.edu/2022/students In this section, we discuss the findings in relation to our research
-and-university/24-academic-regulations) You may not engage in questions. In response to our first research question, our review has
unauthorized collaboration or make use of ChatGPT or other AI found that 23 of the top 50 ranked universities in the THE World Uni­
composition software. versity Rankings 2023 have developed publicly available guidelines for
2. Students must obtain permission from me before using AI composi­ their instructors pertaining to the use of GAI tools in assessment tasks.
tion software (like ChatGPT) for any assignments in this course. That less than half of the universities have accessible guidelines is
Using these tools without my permission puts your academic integ­ concerning given that research suggests that such guidelines have the
rity at risk. potential to raise awareness about academic integrity and reduce aca­
demic misconduct [23,41–43]. Furthermore, scholars have been calling
Along with these statements, the guidelines encouraged instructors for institutions to develop guidelines to support instructors as they adapt
to have open discussions with their students about the use of GAI. These to the GAI world (e.g., [1,19]). Without clear guidelines, instructors may
discussions are suggested to be "open-minded" (CMU), and include dis­ take a defensive approach to GAI and adopt more in-class assessments
cussions about “What will they lose intellectually if they use AI to [7] or feel frustrated as they struggle to adapt their assessment practices
complete their assignments?” (DU), “what ChatGPT or similar AI tools without institutional guidance ([[44]]). It is therefore essential that HEIs
can and cannot do, emphasising those aspects of your assignments that develop GAI guidance and review them regularly as the technology
cannot be outsourced” (Cornell) and what is “permissible use in your develops.
context” (Edin). UCB and UIIC suggest instructors collaborate with li­ Addressing our second research question, the guidelines show a
brarians when discussing with students the impact of using GAI in their concern for the effects GAI tools could have on assessment tasks and
assessments and the value of the process of completing assignments. provide advice on how to address these concerns. There is an
acknowledgement within the guidelines that GAI is here to stay and,
3.3.2. Content of communication therefore, HEIs need to adapt to accommodate it. Our review found that
The university guidelines gave advice to their instructors on the the guidelines included three main areas: the effects of GAI on academic
content of their communications. These include: (1) setting clear ex­ integrity; advice on assessment design; and communication with stu­
pectations; (2) partnering with students to develop class policy; (3) the dents (See Tables 4,5, and 6 for the main themes and sub-themes). The
ethics and limitations of GAI; (4) Importance of originality; and (5) primary aims of the guidelines seem to be to help instructors consider
importance of college learning, intellectual struggle, and process. Due to what might constitute improper use of GAI, develop procedures for
the complexity of GAI tools and their wide range of capabilities, some students on how to acknowledge or cite GAI use, re-design their as­
guidelines suggested that instructors must make it clear if and how GAI sessments so that it becomes more difficult for students to use GAI solely
tools can be used. For example, CU stated, “it is important to be explicit to complete them and discuss if and how GAI can be used in assignment
with students about the expectations around the usage of ChatGPT and tasks. The findings can increase instructors’ awareness of the risks to
other AI tools in your course. For example, ChatGPT is capable of their assessment practices and offer suggestions for how to adapt their
reading a student’s essay and providing meaningful feedback that can practices for the GAI world. They also provide a good starting point for
then be used by the student to make edits. As the instructor, it’s instructors and other HEIs who wish to develop or revise their own
important to be clear about these expectations: can students use the tool guidelines. HEIs can see if their guidelines cover the same areas and
for feedback on their own writing? If so, how should they disclose their align with the top universities in the world.
use? As with all course policies, especially those around academic Specifically regarding assessment design, the guidelines advice
integrity, it is essential for instructors to be explicit and transparent with teachers to design assessments that require creativity and critical
their expectations, and to have frank conversations with their students”. thinking, incorporate contextual elements by connecting course content
CU also suggested that instructors partner with their students to to real-life experiences, implement authentic assessments that allow
discuss acceptable and unacceptable use of GAI. Cornell provided in­ students to apply concepts in real-world situations, focus on the process
structors with three appeals- logical appeal, emotional appeal and per­ and stages of assessment and provide alternative modes of representa­
sonal appeal - to help instructors consider ways to communicate the tion beyond text, such as images, slides, discussions, videos, and audio
issues of using GAI in assessments. These are designed to raise students’ recordings. These suggestions accord with prevailing research and
awareness of the implications of using GAI inappropriately themselves suggestions [7,29,32]. However, the recommendation that teachers
and the community and the value of their degree. Many guidelines should “test” their assessment task by checking how well it can be
emphasised the need to discuss the limitations of AI with students to accomplished solely by GAI and to have students use GAI as part of the
raise their AI literacy. For instance, UCLA suggested that instructors assessment process deserve mention. We believe it offers a principled

8
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

way of integrating the use of GAI in assessment. Table 7

However, questions still remain. The emphasis in the guidelines still A Preliminary Conceptualization of GAI Assessment Literacy.
seem to focus on limiting or preventing GAI use in assessment tasks. Able to recognise the implications of GAI for academic and assessment integrity
Even when suggestions are made on how instructors can integrate them, Able to design assessment tasks that provide space for students to demonstrate
the communications with students focus on the negative implications of learning while incorporating GAI tools in the assessment process
Able to communicate with students about the productive, responsible and ethical
their use. The goal seems to be to design assessments that can elicit
use of GAI in assessment tasks
evidence of learning through responses that cannot be fully auto-
generated with minimal input from students. A noble goal in the short
term. Furthermore, suggestions for proper ways GAI can be used in GAI world. We suspect that any GAI guidelines will need to be reviewed
assessment tasks seem to be limited. This does not align with calls for and updated regularly as the technology and new assessment and
authenticity in assessment tasks [38,45], which relates to using assess­ pedagogical models are developed.
ment tasks that resemble the kind of tasks students have to perform in As stated earlier, the study has several limitations. First, the review
their actual situations, for instance, their current or future jobs [7,20]. only included publicly available guidelines extracted from the reviewed
In future guidelines, a greater emphasis may need to be placed on university websites. The remaining universities may have guidelines
how GAI tools can be made integral to assessment tasks and part of that are not accessible publicly. Second, the review focused on the top 50
students’ lived HEI experiences. Allowing or even requiring students to universities in THE. While this allowed for the review to be done rapidly
use GAI at various stages of the assessment process would, in fact, and systematically, it may not reflect the situation globally. Indeed, the
enhance the authenticity of assessments. Authenticity is concerned with Global North is disproportionately represented. Future studies can
the extent to which the assessment tasks replicate “real world” tasks explore whether the findings here are consistent with guidelines in other
[46]. In the same way as editing tools such as Grammarly are embedded contexts. Third, the review only explored guidelines targeted at in­
in word processing systems, the use of GAI tools is likely to become structors. Universities may have developed guidelines for students on
similarly ubiquitous, hence allowing its use in assessment enhances the GAI use or developed institutional policies governing their use. These
quality of authenticity. guidelines and policies could be reviewed in a future study. Despite
Of course, many of the guidelines rightly point out the limitations of these limitations, the findings offer a rich resource for developing and
these tools and the real concerns their use poses, including bias, inac­ reviewing HEI guidelines and can help inform policymakers, instructors
curacy, privacy and copyright infringement [19,40,47]. Yet, humans and others on the important content of such guidelines.
have always used tools that help them complete tasks, and as new tools
develop, they change the skills and knowledge we need to do certain 5.1. Funding statement
tasks [48]. The jobs of the future will require graduates to be GAI literate
and part of developing these skills will be using GAI tools in productive, This research did not receive any specific grant from funding
responsible and ethical ways in assessments [49]. Understanding how agencies in the public, commercial, or not-for-profit sectors.
we might assess GAI use, such as incorporating it into our rubrics, is also
something we need to consider in future guidelines.
Declaration of Competing Interest
In addition, most of the guidelines focus on one GAI tool - ChatGPT.
This is understandable as it was the first tool to receive widespread
The authors declare that they have no known competing financial
attention. However, there are now a plethora of GAI tools with different
interests or personal relationships that could have appeared to influence
functionalities. GAI tools have been developed that can generate images,
the work reported in this paper.
code, videos, and a wide range of texts [6]. By focusing on one tool
predominantly, it may leave instructors with a limited view of GAI and
Supplementary materials
its effect on assessments. Indeed, once GAI becomes embedded in
common products, such as word processing software, the complexity of
Supplementary material associated with this article can be found, in
considering what are suitable uses of GAI in assessments will only in­
the online version, at doi:10.1016/j.caeo.2023.100151.
crease. Perhaps, having rational and logical conversations with students
about the implications of GAI use, as suggested by Cornell, is the way we
References
need to go. In a sense, adopting trust-based assessment practices that
emphasise the value of the task to student learning beyond grades might [1] Crawford J, Cowling M, Allen K. Leadership is needed for ethical ChatGPT:
be more fruitful. character, assessment, and learning using artificial intelligence (AI). J. Univ.
Clearly GAI has created a highly complex context. In order to navi­ Teach. Learn. Pract. 2023;20(3). https://doi.org/10.53761/1.20.3.02.
[2] Holmes W, Porayska-Pomsta K, Holstein K, Sutherland E, Baker T, Shum SB,
gate these new realities, it has been argued that instructors require AI Santos OC, Rodrigo MT, Cukurova M, Bittencourt II, Koedinger KR. Ethics of AI in
literacies [14]. We argue that the specific nature of assessments means education: towards a community-wide framework. Int. J. Artif. Intell. Educ. 2022;
that instructors need to develop GAI assessment literacy. Based on the 32(3):504–26. https://doi.org/10.1007/s40593-021-00239-1.
[3] Casal JE, Kessler M. Can linguists distinguish between ChatGPT/AI and human
findings above, we provide a preliminary conceptualization of GAI writing?: a study of research ethics and academic publishing. Res. Methods Appl.
assessment literacy (See Table 7). This conceptualization can help act as a Linguist. 2023;2(3). https://doi.org/10.1016/j.rmal.2023.100068.
framework in developing GAI assessment literacy. [4] Sleator L. Almost half of Cambridge students admit they have used ChatGPT. April
21. The Times; 2023. https://www.thetimes.co.uk/article/cambridge-university-st
udents-chatgpt-ai-degree-2023-rnsv7mw7z.
5. Conclusion [5] Nietzel MT. More than half of college students believe using ChatGPT to complete
assignments is cheating. March 20. Forbes; 2023. https://www.forbes.com/sites/m
ichaeltnietzel/2023/03/20/more-than-half-of-college-students-believe-using-ch
This review has shone light on the guidelines adopted by the top atgpt-to-complete-assignments-is-cheating/.
universities around the world pertaining to the use of GAI in assess­ [6] Lim WM, Gunasekara A, Pallant JL, Pallant JI, Pechenkina E. Generative AI and the
ments. The findings highlight the need for instructors to consider the future of education: ragnarök or reformation? A paradoxical perspective from
management educators. Int. J. Manag. Educ. 2023;21(2):100790. https://doi.org/
impact of GAI and develop appropriate assessment practices. As these
10.1016/j.ijme.2023.100790.
guidelines have been developed rapidly due to an emerging technology, [7] Chan CKY. A comprehensive AI policy education framework for university teaching
it is likely that many of the suggestions have not been tested in authentic and learning. Int. J. Educ. Technol. High. Educ. 2023;20(1). https://doi.org/
HEI contexts. As instructors begin to put the suggestions in these 10.1186/s41239-023-00408-3. 38–25.
[8] Warschauer M, Tseng W, Yim S, Webster T, Jacob S, Du Q, Tate T. The affordances
guidelines into practice, we will get a better idea of their efficacy in and contradictions of AI-generated text for second language writers. SSRN
authentic contexts and how assessments will need to be shaped in the Electron. J. 2023. https://doi.org/10.2139/ssrn.4404380.

9
B.L. Moorhouse et al. Computers and Education Open 5 (2023) 100151

[9] Chan CKY, Lee KKW. The AI generation gap: are gen Z students more interested in [29] Cotton DRE, Cotton PA, Reuben Shipway J. Chatting and cheating: ensuring
adopting generative AI such as ChatGPT in teaching and learning than their gen X academic integrity in the era of ChatGPT. Innov. Educ. Teach. Int. 2023. https://
and millennial generation teachers? arXiv.org; 2023. doi.org/10.1080/14703297.2023.2190148.
[10] Chiu TKF. The impact of generative AI (GenAI) on practices, policies and research [30] Perkins M. Academic Integrity considerations of AI Large Language Models in the
direction in education: a case of ChatGPT and Midjourney. Interact. Learn. post-pandemic era: ChatGPT and beyond. Univ. Teach. Learn. Pract. 2023;20(2):
Environ. 2023. https://doi.org/10.1080/10494820.2023.2253861. 07.
[11] Kohnke L, Moorhouse BL, Zou D. ChatGPT for Language Teaching and Learning. [31] Dalalah D, Dalalah OM. The false positives and false negatives of generative AI
RELC Journal 2023;54(2):537–50. https://doi.org/10.1177/00336882231162868. detection tools in education and academic research: the case of ChatGPT. Int. J.
[12] Hwang GJ, Chen NS. Editorial position paper: exploring the potential of generative Manag. Educ. 2023;21(2). https://doi.org/10.1016/j.ijme.2023.100822.
artificial intelligence in education: applications, challenges, and future research [32] Yeo MA. Academic integrity in the age of Artificial Intelligence (AI) authoring
directions. Educ. Technol. Soc. 2023;26(2). https://doi.org/10.30191/ apps. TESOL Journal, 14(3) 2023;716. https://doi.org/10.1002/tesj.716.
ETS.202304_26(2).0014. [33] Phakiti A, Fernandez J, Steinhoff A, Yeo MA, Richards JC. Language assessment: a
[13] Crompton H, Burke D. Artificial intelligence in higher education: the state of the practical approach. Regional Language Centre; 2021.
field. Int. J. Educ. Technol. High. Educ. 2023;20(1):1–22. https://doi.org/ [34] Yang M. New York City Schools ban AI ChatBOT that writes essays and answers
10.1186/s41239-023-00392-8. prompts. January 6. The Guardian; 2023. https://www.theguardian.com/us-news
[14] Ng DTK, Leung JKL, Chu SKW, Qiao MS. Conceptualizing AI literacy: an /2023/jan/06/new-york-city-schools-ban-ai-chatbot-chatgpt.
exploratory review. Comput. Educ. Artif. Intell. 2021;2:100041. https://doi.org/ [35] Yazici A, Yazici S, Erdem MS. Faculty and student perceptions on college cheating:
10.1016/j.caeai.2021.100041. evidence from Turkey. Educ. Stud. 2011;37(2):221–31. https://doi.org/10.1080/
[15] Yang ACM, Chen IYL, Flanagan B, Ogata H. Automatic generation of cloze items for 03055698.2010.506321.
repeated testing to improve reading comprehension. Educ. Technol. Soc. 2021;24 [36] Pope N, Green SK, Johnson RL, Mitchell M. Examining teacher ethical dilemmas in
(3):147–58. https://doi.org/10.30191/ETS.202107_24(3).0011. classroom assessment. Teach. Teach. Educ. 2009;25(5):778–82.
[16] Chu H, Tu Y, Yang K. Roles and research trends of artificial intelligence in higher [37] Pereira D, Flores MA, Niklasson L. Assessment revisited: a review of research in
education: a systematic review of the top 50 most-cited articles. Australas. J. Educ. assessment and evaluation in higher education. Assess. Eval. High. Educ. 2016;41
Technol. 2022;38(3):22–42. https://doi.org/10.14742/ajet.7526. (7):1008–32. https://doi.org/10.1080/02602938.2015.1055233.
[17] Kim C, Bennekin KN. The effectiveness of volition support (VoS) in promoting [38] Sotiriadou P, Logan D, Daly A, Guest R. The role of authentic assessment to
students’ effort regulation and performance in an online mathematics course. Instr. preserve academic integrity and promote skill development and employability.
Sci. 2016;44:359–77. https://doi.org/10.1007/s11251-015-9366-5. Stud. High. Educ. 2020;45(11):2132–48. https://doi.org/10.1080/
[18] Celik I, Dindar M, Muukkonen H, Järvelä S. The promises and challenges of 03075079.2019.1582015 (Dorchester-on-Thames).
artificial intelligence for teachers: a systematic review of research. TechTrends [39] Ambrose SA, Bridges MW, DiPietro M, Lovett MC, Norman MK. How learning
2022;66(4):616–30. https://doi.org/10.1007/s11528-022-00715-y. works: seven research-based principles for smart teaching. 1st Ed. Jossey-Bass;
[19] Rudolph J, Tan S, Tan S. ChatGPT: bullshit spewer or the end of traditional 2010.
assessments in higher education? J. Appl. Learn. Teach. 2023;6(1). https://doi. [40] King MR. A conversation on artificial intelligence, chatbots, and plagiarism in
org/10.37074/jalt.2023.6.1.9. higher education. Cell Mol. Bioeng. 2023;16(1):1–2. https://doi.org/10.1007/
[20] Sullivan M, Kelly A, McLaughlan P. ChatGPT in higher education: considerations s12195-022-00754-8.
for academic integrity and student learning. J. Appl. Learn. Teach. 2023;6(1). [41] Bretag T, Mahmud SW, James C, Green ME, McGowan UP. Core elements of
https://doi.org/10.37074/jalt.2023.6.1.17. exemplary academic integrity policy in Australian higher education. Int. J. Educ.
[21] Cassidy C. Australian universities to return to ‘pen and paper’ exams after students Integr. 2011;7(2):3–12. https://doi.org/10.21913/IJEI.v7i2.759.
caught using AI to write essays. January 10. The Guardian; 2023. https://www.th [42] Möller A. An analysis of university academic integrity policies in New Zealand.
eguardian.com/australia-news/2023/jan/10/universities-to-return-to-pen-and-pa J. Furth. High. Educ. 2023;47(3):338–50. https://doi.org/10.1080/
per-exams-after-students-caught-using-ai-to-write-essays. 0309877X.2022.2130195.
[22] Shepherd T. New AI tools that can write student essays require educators to rethink [43] Tatum HE. Honor codes and academic integrity: three decades of research. J. Coll.
teaching and assessment. May 17. Blog. London School of Economics; 2022. htt Charact. 2022;23(1):32–47. https://doi.org/10.1080/2194587X.2021.2017977.
p://eprints.lse.ac.uk/116271/. [44] Kohnke L, Moorhouse BL, Zou D. Exploring generative artificial intelligence
[23] De Maio C, Dixon K. Promoting academic integrity in institutions of higher preparedness among university language instructors. Comput. Educ.: Artif. 2023;5:
learning: what 30 years of research (1990-2020) in Australasia has taught us. 100156.
J. Coll. Charact. 2022;23(1):6–20. https://doi.org/10.1080/ [45] Darling-Hammond L, Snyder J. Authentic assessment of teaching in context. Teach.
2194587X.2021.2017972. Teach. Educ. 2000;16(5–6):523–45. https://doi.org/10.1016/S0742-051X(00)
[24] UW-Madison Library. Is it a policy, procedure, or guideline? The authors;:2022. 00015-9.
https. doi://development.policy.wisc.edu/2022/06/01/is-it-a-policy-proced [46] Brown HD, Abeywickrama P. Language assessment: principles and classroom
ure-or-guideline/. practices. White Plains, NY: Pearson Education; 2010.
[25] Tertiary Education Quality, Standards Agency (TEQSA). What is academic [47] Moorhouse BL, Wong KM, Li L. Teaching with Technology in the Post-Pandemic
integrity? The author; 2022. https://www.teqsa.gov.au/students/understanding-a Digital Age: Technological Normalisation and AI-Induced Disruptions. RELC
cademic-integrity/what-academic-integrity. Journal 2023;54(2):311–20. https://doi.org/10.1177/00336882231176929.
[26] Janke S, Rudert SC, Petersen Ä, Fritz TM, Daumiller M. Cheating in the wake of [48] Jones RH, Hafner CA. Understanding digital literacies : a practical introduction
COVID-19: how dangerous is ad-hoc online testing for academic integrity? Comput. 2012. https://doi.org/10.4324/9780203095317.
Educ. Open 2021;2:100055. https://doi.org/10.1016/j.caeo.2021.100055. [49] Markauskaite L, Marrone R, Poquet O, Knight S, Martinez-Maldonado R, Howard S,
[27] Von Dran GM, Callahan ES, Taylor HV. Can students’ academic integrity be Tondeur J, De Laat M, Buckingham Shum S, Gašević D, Siemens G. Rethinking the
improved? Attitudes and behaviors before and after implementation of an entwinement between artificial intelligence and human learning: what capabilities
academic integrity policy. Teach. Bus. Ethics 2001;5:35–58. do learners need for a world with AI? Comput. Educ. Artif. Intell. 2022;3:100056.
[28] Pecorari D, Petrić B. Plagiarism in second-language writing. Lang. Teach. 2014;47 https://doi.org/10.1016/j.caeai.2022.100056.
(3):269–302. https://doi.org/10.1017/S0261444814000056.

10

View publication stats