Using Gen-AI

Curtin allows the free use of Gen-AI for study purposes, and you may want to use it for personal or professional use as well. It can help with budgeting, meal planning, personal communication, exercising, job interview preparation, thought organisation, travel plans and more.

However, for assessments, you must have explicit permission from your Unit Coordinator (UC). You should follow any instructions from your UC on how it can be used and ensure that any assignment-based AI use is documented, reviewed and acknowledged.

In brief, when using Gen-AI for assessments, you should:

Gen-AI outputs

Gen-AI outputs are based on probability. Text-based AI tools, known as Large Language Models (LLMs), work like an incredibly advanced predictive text program, generating outputs based on patterns evident in the datasets they’ve been trained on. If an LLM generated the words ‘how are’, the most likely third word would be ‘you’.

While they’re good at generating human-like outputs, they don’t actually think or reason – they’re creating information through complex pattern matching based on their training data. AI outputs need to be carefully evaluated, as they can sometimes generate plausible-sounding but incorrect information (known as hallucinations). Think of it as having a knowledgeable assistant who can help create a first draft but whose work should always be reviewed and refined.

Prompt process

  1. A message (known as a prompt) is sent to the AI.
  2. The AI breaks your message into smaller pieces (called tokens). These can be words, parts of words, or even punctuation marks.
  3. The tokens are contextualised, possibly including information received in earlier prompts, so the AI can begin to form a response based on its training data.
  4. The AI constructs its response to the prompt word-by-word, always selecting the most probable that makes sense in that context. It keeps predicting and adding words until the response is complete.
  5. Once the AI has constructed a complete, coherent response, the answer (or output) is sent back.

As Gen-AI tools can only predict, not reason or think, all outputs need to be independently evaluated for accuracy and suitability. Further, you should edit and improve your responses to fit your specific circumstances and requirements.

Visit the evaluating AI section of this guide for further information.

Before use

Gen-AI tools are trained on large amounts of public data, potentially including your interactions with them. Once personal information has been shared with an AI system, tracking, controlling, or removing that data is very difficult, and it could be shared with other users.

Before using Gen-AI tools for any purpose, consider whether you would be comfortable if the information were shared with a wide audience. If no, ensure you have removed any personal or identifiable information. This includes your name, email address, income details, sensitive legal or medical details, or any other data you wouldn’t share publicly. You can replace personal or private information with general descriptions that remove unnecessary details that don’t need to be shared.

Additionally, ensure that any content you put into the tools does not belong to Curtin University or another author. This includes unit guides, readings, assessment rubrics, and journal articles. These materials are Curtin’s intellectual property, and giving them to AI tools is a breach of copyright.

Before selecting a Gen-AI tool, consider its privacy policy and ask yourself the following questions:

  • Are your prompts being used to train the AI model?
  • How is your data being stored, and can it be deleted if needed?

Permitted use

In general, you can use Gen-AI for study-related purposes to help improve your efficiency and understanding. You don’t need specific permission to use Gen-AI for uses not related to your assessments. If in doubt, you should check with your lecturer.

  • Explaining difficult concepts, theories or maths problems
  • Identifying alternate keywords to assist with searching
  • Creating revision questions and practice problems
  • Roleplaying scenarios, such as professional interaction or potential presentation questions
  • Creating a study schedule or calculating the time tasks will take
  • Breaking down or reinterpreting tasks
  • Interpreting assessment feedback
  • Writing emails
  • Organising group projects
  • Turning a brain dump into actions
  • Troubleshooting programs (e.g., MS Word, the Adobe Suite)

Template

To do any of these tasks, modify the prompt template below to fit your specific task. You may not need to use all elements of the prompt template to get a useful answer.

I am a university student currently studying [discipline] and have [scenario or problem].

Please help me [what you want to achieve] by [task you would like the Gen-AI model to complete].

Include [additional elements, information for context].

Provide your answer in [specific format or tone].

Example #1

I am a university-level student. Please help me understand the relationship between taxation and business taxpayers, breaking down the key elements and providing examples. Include information relating to an Australian context. Provide your answer in a short paragraph using simple language that a high schooler can understand.

Example #2

I am a university student currently studying Occupational Therapy and need practice roleplaying a professional interaction with a potential client. Please help me with this process by pretending to be a patient who is recovering from knee surgery. Provide your answer in a chat format and we will talk back and forth. Please let me know how I did and how I can improve in future.

Example #3

Please help me discover alternative search terms for the following terms keywords: social media, body image, adolescents.

For information on building more in-depth prompts, view the effective prompts section.

If you receive specific approval from your Unit Coordinator, there are various ways Gen-AI can be used in assignments. Again, it is important to stay within the bounds of the permissions granted. Not all suggestions below may be appropriate, even when given permission, so check your unit guidelines carefully. What is allowed in one unit may not be authorised in another, so contact your UC if you are unsure.

In addition to assesses your understanding of a topic or concept, university also builds your ability to form opinions, think critically, and communicate your ideas. For these reasons, explicit permission is required to use Gen-AI in the creation or critical analysis of content, including:

  • Writing any part of your assessment
  • Paraphrasing or re-writing assessment content
  • Searching for sources or providing research to be used in assessments
  • Providing feedback on your assessment before submission
  • Answering or checking online test questions
  • Creating presentations based on unit materials, readings or assessments.
  • Translating written assessments into English. Your English language proficiency is being tested, and the tools are often imperfect.
  • Finding mistakes in math problems, code and calculations
  • Transcribing speech-to-text

If you have permission to use Gen-AI tools for any of the above tasks, review the prompt engineering information below to see which formulas will be the most effective for your purpose.

Uploading Curtin-owned content like unit guides, assessment tasks and questions, or academic resources breaches Curtin’s copyright and intellectual property. Therefore, you need to avoid:

  • Putting unit content (e.g. slides, unit outlines, speech-to-text) into text-based Gen-AI to create summaries.
  • Uploading any copyrighted content, including unit materials, scholarly journal articles, other texts etc.

Additional examples of appropriate and inappropriate use of Gen-AI software can be found in Curtin’s Academic Integrity Guide for Students.

Effective prompts

Useful outputs from Gen-AI technology rely on submitting a detailed request or piece of text, also known as a prompt. Prompt engineering is the process of designing a prompt that will produce the best outcome. Some of the interactions you have with AI will be conversation and you won’t need to build a structured prompt. However, if you want a response in a specific format or a greater level of detail, you will need to use a prompt engineering technique.

There are a range of prompt engineering techniques to help you get the most effective results:

Few-shot prompting provides the AI with examples to help it better understand the task and what you expect to receive as a result. The AI model learns from your previous messages and the guides you provide it with.

By providing two or more examples, the AI can recognise patterns and provide more accurate responses.

Example prompt

Classify the following symptoms into one of the categories: Common Cold, Flu, or Allergies. Here are some examples:

Examples:

  1. “I have a runny nose, sneezing, and itchy eyes.” = Allergies
  2. “I’ve had a fever, body aches, and chills for two days.” = Flu
  3. “My throat is sore, and I have a mild cough but no fever.” = Common Cold

New input: “I have a fever, fatigue, and a headache.”

Expected Gen-AI response:

Based on your new input, here’s the classification:

“I have a fever, fatigue, and a headache.” = Flu

These symptoms are commonly associated with the flu, especially when they occur together.

When to use this technique: When you have a mixed range of information that needs to be sorted into different user-determined categories.

COT prompting is when the AI model is encouraged to explain its reasoning step-by-step before providing an answer. This method allows you to see the steps taken to reach an answer.

Add the phrase ‘Let’s think step-by-step, including explanations’ to the input and the model will break down its response to include details about the process.

Example prompt

Balance the following chemical equation. Let’s work it through step-by-step, including explanations about why each step is done.

Balance the reaction: Al + HCl → AlCl3 + H2

When to use this technique: Reasoning tasks like maths or science problems, decision-making, and logical analysis.

Prompt chaining encourages the AI to break down complex tasks into a series of smaller, more manageable prompts. Following the first prompt, each additional prompt in the chain provides further information to help the result reach its final desired state. It works similar to an ongoing conversation with the Gen-AI tool.

Example prompt

Scenario: Create a comprehensive semester schedule that includes assignment due dates, class schedules, holidays, personal events, and work shifts.

  • Prompt #1: Here is a list of key dates, class times, and events for the semester. Organise them into categories: Classes, Assignments, Holidays, Personal Events, and Work Shifts. INFORMATION: [Include details and dates for each]
  • Prompt #2: Now, create a weekly schedule with time blocks, filling in classes, work shifts, holidays, and personal events. Make sure to leave time for study and personal activities.
  • Prompt #3: Place the assignment due dates, holidays, and personal events onto the calendar, making sure to leave time for preparation and recovery.
  • Prompt #4: Now, identify any gaps in your schedule where you can allocate study and revision time for exams and assignments.

When to use this technique: Breaking down large tasks, generating structured content, complex work requiring multiple steps

CARE stands for Context-Action-Response-Example. In your prompt, you will provide details under headings linked to each letter in the acronym.

  • CONTEXT: Describe the situation, role, experience level, location, and the project and/or task you are working on
  • ASK: Request a specific action from the AI, including the role you want it to take on, what output you want, the steps you want it to follow
  • RULES: Provide constraints, including word or sentence limits, writing style requirements, desired format, etc.
  • EXAMPLE: Provide examples of what you do or don’t want the model to produce, including good and bad examples. Sometimes, you may not need to provide examples to the model.

Example prompt

CONTEXT
I am a medical student at university and need to increase my experience interacting with patients.

ASK
Please act as a patient who has arrived at a GP clinic presenting with an illness. You will decide on this illness before our conversation but will not tell me what this is. We will roleplay a conversation mimicking a doctor’s appointment, where I will ask you questions about your symptoms.

RULES
Please provide all answers in a chat format as though we are having a conversation. Use an informal tone. You will not tell me whether I am correct until I provide a diagnosis following the phrase “OFFICIAL DIAGNOSIS”.

Let me know if you have understood this. Do you have any additional pieces of context or rules that will benefit my practice here? Let me know when you are ready, and we can begin.

When to use this technique: Roleplay scenarios, acting as an academic tutor to explain difficult concepts or problems, etc.