Generative Artificial Intelligence

Generative Artificial Intelligence

What is Generative Artificial Intelligence?

Generative Artificial Intelligence (GenAI) are digital tools that have ‘learnt’ from large resource databases (such as sections of the internet), which can generate content based on what it has ‘seen’ before in its development phase. Often these tools continue to ‘learn’ as users respond, positively or negatively, to their outputs.

There can be a variety of inputs and outputs involved for different GenAI tools, from text, images, sounds, animation, 3D models (Nvidia, n.d.), computer code, etc. Content can be small editing corrections (like spelling, grammar and auto-complete recommendations) to whole essays or pictures. These can be generated very quickly, and where tools offer paid subscription options, a user might generate a lot of content in a short amount of time. These tools frequently offer a chat-like interface, making ‘inputs’ (questions or ‘prompts’ from users) feel conversational, more comfortable than hunting for information online or in a book or journal article.

GenAI can be a wonderful tool, and is increasingly being incorporated into business practice and tools. But it has limitations, and deeper than this, it is critical for learners not to sidestep their own learning journey by over-reliance on such a tool. These guidelines are intended to assist learners navigate the limitations and dangers of these tools at a broad level, in part to meet the assessment regulations of their assignments, but also for their practice more generally.

The College's Policy Statement regarding Generative Artificial Intelligence is available under the Quality Assurance Handbook Part B Section 3.

Table of Contents:

• Key GenAI Tool Points to Consider
• Why be worried about GenAI?
• Do's and Don't's of using GenAI
• Advice for Prompts in GenAI Tools
• Further help with GenAI Tools
• Guide References

 

Key GenAI Tool Points to Consider

Where it is permissible to use GenAI tools in the preparation for or completion of an assignment, consider these final key points (adapted from Griffith University, 2024) before utilising any GenAI output:

• AUTHORISATION - are you certain that you are allowed to use GenAI in your assignment?
• CRITICAL THINKING SKILLS - any outputs MUST be evaluated for accuracy, relevance, and bias.
• BIAS - GenAI is trained on information from human-generated datasets, so any user MUST be wary of potential biases like representations or assumptions of gender, race, or cultural difference in the outputs.
• SOURCE VERIFICATION - GenAI tools can hallucinate or make up sources and references, often very convincing ones. Any source cited should be checked (see the Library’s guides on searching for sources), and verified that the source actually says what the GenAI outputs presented it as saying.
• PRIVACY AND DATA SECURITY - GenAI tools are third-party software platforms and tools, which collect and store data you provide, including the questions you ask and information you provide (assignment tasks and questions must NEVER be submitted to a GenAI tool, just as in a business, you would be forbidden from submitting commercially sensitive or personal data to an external system)
• REFERENCE AND DECLARE USE OF GENAI - you should always cite GenAI outputs where they are used in an assignment, and follow the referencing guidelines. Failure to declare GenAI use can result in investigations and penalties for academic misconduct.

KEY GUIDANCE - treat any GenAI tool as an unreliable assistant - it can help in some tasks, but should not be left to do the core work, and cannot be trusted to always get it right, so must be used cautiously.

 

Why be worried about GenAI?

GenAI can make mistakes

GenAI tools, despite their name, are not intelligent - like autocomplete tools which try to predict the next word in the sequence (based on what it’s ‘learnt’ from its training database), it tries to render what is the most likely ‘answer’ or output in response to a prompt (ETBI, 2024). This means they do not ‘know’ the correct answer (it just replicates what it’s been trained to see as the most common answer), they can hallucinate or make up information (following the ‘pattern’ of other similar information it trained on), and they can perpetuate biases and stereotypes since the material it’s been trained on carry unintentional assumptions or frameworks, like American and European media where certain principles, beliefs, racial or gendered frameworks, or other cultural norms dominate (Brown et al., 2020 p.10).

GenAI does not always help you learn

Beyond these immediate limitations and weaknesses, GenAI tools can risk a learner’s learning process. Because the tools can feel conversational, comfortable, and are so quick, they feel easier to use to find information - but properly learning something, absorbing the information, and synthesising it into a new understanding, is rarely quick and comfortable. If it was, there would be little value in the parchment that evidences a learner’s achievement in completing a degree. This is captured by the concept of ‘Desirable Difficulty’ (Bjork, 1994) - for learning to really ‘sink in’, it has to be worked through, engaged with, struggled with a bit. ‘Easy’ information retrieval does not lead to long-term memorisation or understanding. Further to this, learners still need the underlying knowledge and understanding to identify when the outputs from a GenAI tool are incorrect, hallucinated, or perpetuating biases - this emerging ‘AI literacy’ is, in addition to the traditional ‘information literacy’ and generic digital and media literacy, needed by the modern member of society (Bundy, 2004). Learners still need to be able to explain or defend any work they submit for an assignment, and where they cannot, this will incur a fail grade, or a finding of academic misconduct.

Academic Integrity

Misuse of GenAI undermines a learner’s learning process, but also invalidates their assessment. Assignments are not arbitrary trials or challenges to make students’ lives difficult, they are to evaluate the degree of learning achieved. For formative assessments, this evaluation helps a lecturer identify what areas need more support or attention, and their feedback helps a learner improve for their final assessments. For summative assessments (anything that carries a grade towards an overall module mark), this evaluation allows the lecturer and College to certify that the skills and knowledge expected from a graduate have been achieved - it is this certification that makes a parchment or transcript meaningful, it is a declaration to future employers and society at large that a learner has a certain level of competence and knowledge. Overreliance or the misuse of GenAI, like any other academic misconduct, undermines that certification function, it diminishes the trust employers and society can have in an award or even the College as a whole, and threatens not just a learner’s own achievements, but also that of classmates and future learners. This is why the College takes academic integrity and the appropriate use of GenAI tools so seriously.

Other Ethical considerations

Further to putting a learner’s learning process at risk, or invalidating an assessment, or perpetuating harmful biases and stereotypes, GenAI tools can be of concern in their use of data, and their sustainability implications.

Many GenAI tools require users to subscribe or register, providing their personal information to these software companies in the process. While some try to be explicit about their use of their user’s information and activity, others are less transparent, and learners should be cautious of signing up to any third-party system, in case their data is misused. As the GenAI tools often continue to ‘learn’ through their interactions with users (by collecting positive or negative feedback on the effectiveness of their outputs, such as follow-up prompts), user activity may be recorded and stored for use without explicit or intentional authorisation from that user. Even before these tools monitor user activity, the data originally scraped to build a ‘learning database’ may include material with copyright or other proprietary restrictions, and have nevertheless been used without permission. Inputting content into a GenAI tool that might be commercially sensitive may be a huge risk, and many organisations have already set out policies restricting this by employees.

GenAI tools rely on immense computing power, which in itself has implications for environmental sustainability, with significant resources poured into maintaining and running the data centres answering a user’s queries. While the least visible implication for a user generating assignment tips or an image to accompany a presentation, such environmental and societal implications should be borne in mind for all users.

While there are a lot of considerations to take into account when using GenAI, both practical and ethical, it is increasingly becoming incorporated into the world of work, so the College recommends that, when engaging with these tools, all learners and users do so critically.

 

Do's and Don't's of using GenAI

GenAI tools should be used for supporting and enhancing learning, for practice and inspiration, but it should not be used in place of completing an assignment or exam that is assessing knowledge, skill or understanding - these tools should enhance a learner’s skills, not bypass the learning process (ETBI, 2024).Consider some Do’s and Don’ts of using GenAI tools, in different stages of a learning process:

GenAI can help with:	GenAI should NOT be used as:
Introductions to new topics/ ideas Like Wikipedia or other internet searches, GenAI can help set out introductory explanations or breakdowns of a new topic or idea, but like Wikipedia, it should be treated cautiously, and ideally as a springboard to progress deeper to more trustworthy sources.	An answer-generator GenAI outputs should never be used without critical evaluation in an assignment, and NEVER in a test, MCQ or exam; Even in Open Book exams, learners should be cautious what resources they utilise, e.g. some search engines have incorporated GenAI outputs as part of their results - this still constitutes drawing on GenAI during an exam.
Brainstorming and summarising Listing key ideas and topics; Suggesting an assignment outline; Explaining complex concepts in simpler terms.	Alternative to Studying Generating answers from GenAI should NEVER be confused with active studying - proper revision and practice should be exercised by learners.
Proofreading GenAI can look over your work and make suggestions, but you may NOT let it write for you (unless that is the purpose of the assignment - a lecturer will make this explicit in the assignment). Learners should be careful as GenAI tools may by default try to rewrite a paragraph even if not instructed to.	Rewording tool Where GenAI is used to proofread or offer suggestions to any work you have written, you MUST NOT copy and paste any rewritten outputs, as this is no longer your words and would constitute plagiarism.
Research help Help to produce search prompts for finding articles in major databases like EBSCO; Help with formatting references to a particular style.	Research-replacement GenAI cannot be trusted to provide real sources or genuine quotes, every output should be reviewed and checked; GenAI may perpetuate biases and miss key research points.
Assignment practise Quickly provide ‘model answers’ you can practise analysing and critiquing; Provide quizzes and flash cards to revise and practise with.	Assignment-generator Outputs from a GenAI tool should NEVER be used in an assignment where it is not authorised to use GenAI (see below), and where it is authorised to use GenAI, any outputs MUST be identified and clearly noted and referenced.
Practice as a debate partner Digging into ideas and having an educated ‘conversation’ with GenAI on the topic; Practice for a QnA session.	A tutor to share assignment questions with Learners should NEVER share assessment questions or instructions with GenAI platforms, sharing assessment tasks with third parties, including GenAI tools, is against College policy and constitutes misconduct.

 

Using GenAI in Assignments

The lecturer should make clear in the Assignment Brief what level of GenAI, if any, can be used - the following table (adapted from Perkins, Furze, Roe and MacVaugh, 2024) in the instructions will mark what usage is allowed:

1	2	3	4	5
NO GEN-AI	GEN-AI-ASSISTED IDEA GENERATION AND STRUCTURING	GEN-AI-ASSISTED EDITING	GEN-AI TASK COMPLETION, HUMAN EVALUATION	FULL GEN-AI
The assessment is completed entirely without GenAI assistance. This level ensures that students rely solely on their knowledge, understanding, and skills. GenAI must not be used at any point during the assessment.	GenAI can be used in the assessment for brainstorming, creating structures, and generating ideas for improving work. No GenAI content is allowed in the final submission.	GenAI can be used to make improvements to the clarity or quality of student created work to improve the final output, but no new content can be created using GenAI. GenAI can be used, but your original work with no GenAI content must be provided in an appendix.	GenAI is used to complete certain elements of the task, with students providing discussion or commentary on the AI-generated content. This level requires critical engagement with GenAI generated content and evaluating its output. You will use GenAI to complete specified tasks in your assessment. Any GenAI created content must be cited.	GenAI should be used as a ‘co-pilot’ in order to meet the requirements of the assessment, allowing for a collaborative approach with GenAI and enhancing creativity. You may use GenAI throughout your assessment to support your own work and do not have to specify which content is GenAI generated.

 

• Save a copy of any prompts used and outputs generated (you may include this as an appendix in your assignment) - if the assignment is looking to assess the process through which your work was developed, this appendix can demonstrate your process. If your submission is suspected for improper use of GenAI tools, this appendix may demonstrate what material was used or considered, and assist the Academic Impropriety Committee in fully understanding your case.
  • BEST PRACTICE TIP: save a copy of each output in a folder on your Drive or PC (some GenAI tools offer only a limited ‘memory’/ record of what you asked and generated).
• Review any material generated - accuracy is not guaranteed, and false or hallucinated references occur (see above for further info).
• Include a statement in your Assignment Cover Page detailing how GenAI was used.
• Reference the GenAI tool where you’ve used it throughout your assignment.
• Integrate the answer, if appropriate to use (e.g. option 3 above), into your assignment in the same way you’d integrate any external source, with referencing (see the referencing guide below).

 

Using GenAI in Exams

Unless explicitly stated otherwise, GenAI is NOT to be used in exams, MCQs or tests.

In studying or revising for an exam, learners must not mistake generating answers from a GenAI tool for active studying - quick answers rarely ‘sink in’ and will not be effectively retained for the actual exam.

Learners undertaking Open Book Exams must be cautious of any resources they use - some search engines have integrated GenAI-generated ‘results’ into their responses to queries. This may not be easily noticed when moving quickly through sources during a timed exam, but reference to GenAI outputs and tools during an exam will be considered misconduct. Open Book Exams do not allow lifting or transcribing material from study notes, although key quotes or terms can be referenced, but learners must be careful of drawing from their notes during an Exam, where they have generated those notes through GenAI tools.

Referencing GenAI

The College Library offers guidance on referencing GenAI tools, in line with the developing referencing guidelines of the MLA and Chicago referencing standards, but at the very least, where GenAI content has been integrated into an assignment, the following should be in place:

• In-Text Citations: e.g. OpenAI (2023); (OpenAI, 2023)
• Reference list:  Company. Year. Name of Generative AI (Version number if known). [Generative AI]. [Date accessed]. Available from: URL of specific output if available.  
• Appendix of Prompts/ Outputs

 

Advice for Prompts in GenAI Tools

Where it is permissible to use GenAI tools in the preparation for or completion of an assignment, consider these guidelines (adapted from Griffith University, 2024) to best utilise the tools:

• Clearly state the question or topic, be concise and specific
  • Avoid vague or ambiguous language.
  • The more precise a prompt is, the better GenAI can provide relevant and accurate outputs.
  • Consider indicating what audience the output is intended for (e.g. a managers meeting vs a general public announcement), what the purpose of the output will be (e.g. educational training vs advertising vs a brief summary), and the context (e.g. a training course vs a sales bid).
• Complex topics can be broken down into smaller units to focus the outputs to particular areas.
• The style, format or level of depth can be specified, e.g. “in-depth analysis” vs “bullet-point list”, or specifying a word count.
• Directing the GenAI tool to adopt a ‘persona’ or reply in a specific style or tone can result in more tailored, contextualised responses (e.g. a teacher vs a health official vs a food blogger, etc.).
• Most GenAI tools offer the functionality to follow up a response with additional questions to dig further, refine or redirect to an intended answer or output, like a conversation - asking the tool to clarify, elaborate, provide examples, consider alternative arguments, or revise an answer to be simpler, can refine the outputs.
• Where GenAI is being used to generate images or non-text media, be explicit about:
  • Format - consider images, patterns, banners for PowerPoint or posters, screen or social media post backgrounds, animations, graphs, charts, etc.
    • You may be able to specify the outputs orientation (landscape or portrait) and aspect ratio
  • Specifics - be explicit about what is in the image, and if it is an animal or person, what action is it taking (e.g. sitting, dancing, running, etc.) and where should it be situated in the image and in relation to other objects/ animals/ people/ background details.
  • Colours and patterns - be specific about which colours should be included (consider the tone, purpose and mood they convey, and cultural or brand associations)
  • Stylistic elements - be wary of copyright infringements (e.g. asking GenAI to replicate a specific artist), but consider styles, e.g. abstract, cartoon, pop art, concept art.
• The image-prompt formula should include SUBJECT+ DETAILS + STYLE + FORMAT
• Where GenAI is being used to assist in text editing and proofreading, to avoid it rewriting your answers (and therefore no longer being in your words, and potentially breaching academic integrity), consider incorporating explicit instructions like the following into the prompt:
  • “...do not rewrite”
  • “In a bullet point list…”
  • “Look for unnecessary adjectives and adverbs that can be removed without changing the meaning”
  • “Edit this text by only removing words and adding punctuation”
  • “Identify incidences of long or wordy sentences”
  • “Identify any grammar or punctuation errors”
  • “Offer suggestions to improve organisation or structure of paragraph”
  • “Suggest words that could improve the formality of writing”
  • “Identify repetitive or redundant language”

 

Further help with GenAI Tools

If you have questions about whether GenAI tools can be used in your assignment, or how much it can be used, or if there are specific tools that you can or cannot use, speak to your lecturer.

If you have general questions about using GenAI in assignments, such as appropriate referencing practice and over-reliance on GenAI outputs, contact the College Library, who host workshops on this and referencing generally.

 

Guide References

Bjork, R. A. (1994). ‘Institutional impediments to effective training’. In D. Druckman and R. A.Bjork (Eds.), Learning, remembering, believing: Enhancing human performance (pp.295-306). Washington, DC: National Academy Press

Brown, T.B. et al. (2020) ‘Language Models are Few-Shot Learners’, 34th Conference on Neural Information Processing Systems, Vancouver, Canada. Available at: https://proceedings.neurips.cc/paper/2020/file/1457c0d6bfcb4967418bfb8ac142f64a-Paper.pdf (Accessed 08 August 2024) 

Bundy, A. (2004). Australian and New Zealand information literacy framework: Principles, standards and practice (2nd ed). Australian and New Zealand Institute for Information Literacy (ANZIIL) and Council of Australian University Librarians (CAUL).

ETBI (2024) ‘Avoiding Plagiarism: ChatGPT and Generative AI’. Available at: https://library.etbi.ie/plagiarism/AI (Accessed 08 August 2024).

Griffith University (2024) Using Generative AI ethically & responsibly, available at https://griffith.h5p.com/content/1292031535876431169#h5pbookid=1292031535876431169&chapter=h5p-interactive-book-chapter-c84b0171-0d39-4c21-b4d8-9d29976fa487&section=0 (Accessed 08 August 2024)

NVIDIA (no date) ‘What is Generative Ai?’. Available at: https://www.nvidia.com/en-us/glossary/generative-ai/ (Accessed: 22 July 2024).

Perkins, M., Furze, L., Roe, J. and MacVaugh, J. (2024) ‘The Artificial Intelligence Assessment Scale (AIAS): A Framework for Ethical Integration of Generative AI in Educational Assessment’. (2024). Journal of University Teaching and Learning Practice, 21(06). https://doi.org/10.53761/q3azde36