Teaching about and teaching with AI

In this section, we consider some of the opportunities for learning and teaching. The following examples might be used directly in teaching or to support students to study with AI.

Mollick & Mollick (2023) identify seven approaches for integrating AI in classrooms that help expand considerations of the potentials of AI:

• Mentor: where the AI provides feedback, e.g., students get feedback on the structure of an essay.
• Tutor: where the AI gives direct instruction, e.g., chatbots generate explanations for concepts and ask open-ended questions.
• Coach: where the AI prompts metacognition, e.g., students use a chatbot to reflect on their experience working on a group project.
• Team Mate: where the AI increases team performance, e.g., providing differing perspectives or critiques of the team's ideas.
• Student: where the AI receives instruction, e.g., chatbot acts as a novice learner and asks your students questions about a topic.
• Simulator: where the AI provides deliberate practice, e.g., chatbot role-plays as a patient or client in a relevant scenario.
• Tool: where the AI is used to accomplish tasks, e.g., brainstorming, grammar checking, data cleansing, scenario building.

For examples of AI in teaching from across Monash that fit within the seven approaches outlined above, explore Be Inspired and the AI in Practice Professional Learning Hub (Monash staff only).

Expand the accordions for a few more activities to consider integrating into teaching.

Feedback can come from multiple sources, including teachers, classmates, peers, the physical or social environment, and now AI. Carless and Boud (2018) champion the need for feedback literacy and developing capacities to appreciate feedback, make judgements, manage affect and take action.

The process of providing feedback can be enhanced through responsible AI integration. Educators and peers can use generative AI tools to edit and refine feedback to align with the principles of effective feedback. Additionally, large language models (LLMs) can be prompted to generate feedback directly for students. Henderson et al. (2025) explored student perceptions of AI-generated feedback, highlighting both potential benefits and challenges.

Select the tabs to learn about AI can assist in providing feedback

Additional considerations when using AI for Feedback

Just as we expect students to engage meaningfully with AI if they use it to produce evidence of their learning, we expect the same from educators if they use AI to evaluate and provide feedback on that evidence.

Feedback is broadly understood as inputs to help further improve work and is distinct from, but intertwined with, marking evaluations of the quality of work. There is more scope to integrate AI into feedback processes and more constraints on using AI in marking which has implications for course progression.

Expand the accordions to learn more

Marking represents judgement of quality based on expressed criteria and provides indicators of achievement of learning outcomes.

In evaluating the achievement of learning outcomes, we need to carefully determine what knowledge and skills need to be demonstrated independently by an individual human and which knowledge and skills can be demonstrated collaboratively with multiple intelligences (with classmates or with AI assistance). Similarly, for marking, we need to consider where human judgement is required and where AI can assist. For example, can AI perform spelling, grammar, readability and citation checks while humans perform a check of evidence and argument in a collaborative marking process?

The following are some key considerations as approaches and practices regarding responsible use of AI and marking evolve:

• Privacy rights must be maintained. Monash provided and supported enterprise tools have privacy protection for Monash data. Do not upload any personal information that does not have adequate privacy protection and contributes data to public training of AI models. For example, do not enter student work into an AI platform without the student’s explicit permission. Anonymising is a further crucial strategy to protect privacy.
• Intellectual Property (IP) rights need to be maintained. Students maintain some ownership of their work and we absolutely should not upload student work to AI systems that do not have adequate privacy protection and will be contributing the student work to public training of AI models.
• Accuracy is crucial and we must have confidence that the AI system providing marking advice or marking is generating accurate responses. We need to be able to explain how the AI is making judgements and how we are checking and confirming judgements, maintaining a “human in the loop” for marking with AI processes.
• Transparency is required and we must inform students how AI will be used in marking processes.
• Utility, what parts and aspects of an assessment can AI productively contribute to whether judging criteria or offering feedback?
• Personalisation, how can AI contribute to tailored responses calibrated to our diverse cohorts of students? If AI is used in marking then how can we create human connections that are a backbone of rich university experiences?

Integrating AI tools with traditional study strategies can enhance educational outcomes without compromising foundational skills.

Mollick & Mollick (2023) approaches for integrating AI in classrooms can also be applied as effective study strategies. Some examples include:

• Create practice quizzes, multiple choice and short answer questions with feedback
• Summarise information and generate note-taking frameworks to scaffold complex readings
• Scaffold stages of the writing process including brainstorming and refining research questions, as well as providing feedback on written structure, clarity and flow of drafts

Learn HQ has information for students about ai-assisted study strategies:

• Writing with generative artificial intelligence
• Effectively summarising with generative artificial intelligence

Distinguishing between categories of tools can help educators to consider in which contexts they might choose to use particular kinds of AI tools for teaching:

Narrow vs. General AI

What kind of jobs does the AI do? Narrow AI focus on accomplishing a specific task or problem (e.g. play checkers or another game, search engine, speech recognition, etc.) and cannot understand or apply knowledge beyond its domain or scope. While general AI can accomplish a range of tasks (e.g. Google Gemini or OpenAI ChatGPT being multimodal and able to recognise and create text, image, code, etc.) but cannot do everything.

Stand-alone vs. Integrated AI

Where does the AI operate? Stand-alone AI run as independent systems (e.g. Anthropic’s Claude or OpenAI ChatGPT LLM) but AI can also be integrated into other software and platforms (e.g. OpenAI ChatGPT LLM powering Copilot assistance in Microsoft’s Word, Excel, PowerPoint, Bing etc. or facilitating the processing and responses for many rapidly emerging and evolving AI tools).

Open source vs. Closed source

How is the AI distributed? Open source approaches involve making the source code, data, and documentation of an AI model or platform publicly available and accessible for anyone to use, modify, and share. Closed source or commercial approaches involve keeping the source code, data, and documentation of an AI model or platform private and proprietary, and only allowing authorised users or customers to access and use them.

Even though ChatGPT, Claude and Gemini have drawn a flurry of attention to AI, there are a number of AI based tools already shaping our lives and practices, and new tools emerging daily. This section offers some guidance on deciding which tools to incorporate into learning & teaching.

University provided AI tools

The university provides a number of digital tools that have enterprise agreements to protect the data used and the security of staff. Monash data should only be used in systems and tools provided to staff. Refer to AI at Monash for more information including the Gen AI services currently offered at Monash including Google Gemini, NotebookLM and Copilot.

Approaches to selecting AI tools

There are many kinds of tools available, but ultimately your choice of tools should be driven by safety and pedagogical considerations, i.e. what do you need students to be able to learn or do?

Start with deciding on the use case, for example:

• I want students to get assistance understanding readings/videos prior to interactive class meetings.
• I want students to get assistance searching for source materials for a research project.
• I want students to get assistance with creating presentation materials.

Once you are clear on your use case, search for relevant tools to support the activity.

Like choices of word processor, web browser or search engine, sometimes there are families of tools that have different interfaces and functions. You may want to “test drive” a couple of tools in a family (eg. video summary tools, essay explanation tools, tutoring tools, etc.) or seek out reviews and comparisons of the tools to assist with determining which best suits your context and your intended use case.

The university is increasingly adding new AI tools to our toolbox. See the Artificial Intelligence at Monash website for updates on access and developments.

Considerations for the responsible use of AI tools in teaching

As educators embrace AI tools to enhance teaching and learning experiences, it's paramount to consider the responsible use of such technology. Refer to Responsible use of AI in education for key considerations to ensure ethical, equitable, and effective integration of AI tools in educational settings.

The resources supporting safe use of AI page provides guidance to navigate through decisions to ensure safe and responsible use of AI tools and systems.

Making responsible choices

Prompt engineering is the art of crafting questions or prompts that guide an AI model, like a chatbot, to generate useful and relevant responses. It is similar to giving clear directions to someone to help them understand exactly what you information want. The ways you ask and the context included in the request shape the quality of the responses. If the response is not meeting your needs then expand, clarify or reformulate the request.

Tips for good prompt engineering

1. Be specific: Avoid ambiguity or multiple interpretations of your prompt
2. Use keywords: Similar to using a search engine, keywords are what guide the AI’s response
3. Keep it concise: A short, clear prompt is more likely to yield useful results
4. Provide context: Help the AI to understand the situation to improve relevance
5. Use proper spelling and grammar: AI can handle minor errors, but incorrect language can lead to ambiguity
6. Experiment with different formats: Choose a prompt structure that makes sense for your goal (e.g.) direct, Yes/no, or open questions, commands, MCQ, comparisons, fill-in-the-blank(s), hypothetical scenarios and problems to solve.
7. Be mindful of bias: Avoid emotive language and phrasing that could introduce bias to the response. Remember that the AI is a people-pleaser! It’s looking for the ‘best’ response, i.e. the one that stops you from re-prompting
8. Understand the limits: AI is a model. It can be wrong! Use your judgement to analyse the response and seek additional / corroborative sources if necessary
9. Iterate and refine: Work with the AI. Give it feedback on its responses and tell it what is wrong with the output. Probe the response, rephrase or refine your prompts to get more detail or a different response

Prompt types

There are different ways to prompt, they yield different outcomes and reflect different levels of engagement when working with an AI.

For further descriptions with examples see Google's Prompt Engineering for Generative AI resource.

Prompt strategies

There are many ways for formulating prompts, including ways to set the context and guide the outcomes of requests.

The basic components of a prompt include specifying what you want/want to do (required), providing contextual information (optional), giving system or style instructions (optional) and offering examples (optional).

Google’s PARTS heuristic (PDF) offers one way of conceptualising the key parts of a prompt:

• Persona: Identify your role
• Aim: State your objective
• Recipients: Specify the audience
• Theme: Describe the style, tone, and any related parameters
• Structure: Note the desired format of the output

For further information here is a short video simple guide to effective prompt writing (9 minutes) produced by an academic through the Thinking in public outlet.

Adobe offers related advice for crafting image prompts.

See related prompting advice to students that has been developed for LearnHQ, Creating effective prompts when using artificial intelligence.

Prompt libraries

Writing prompts takes practice and is always an iterative process of testing and refining. However there are a number of useful prompt libraries sharing developed prompts that can be modified for your given purposes.

• Ethan & Lilach Mollick: Prompt Library
• Anthropic Claude Prompt Library
• Prompts for H5P content creation

Using AI responsibly includes always openly acknowledging and explaining where and how AI has been used. Educators should model desirable behaviours by being transparent about their AI use.

For more details, see AI acknowledgement.

This suite of short videos in LearnHQ also help explain the importance of Ai acknowledgement and transparently explaining how things are produced.