Short-term approaches to mitigation and AI literacy

Start a conversation

Talk with your UTS course director, subject coordinator, head of faculty and students about the use of large language models and other generative AI. Send an announcement in Canvas to your students to encourage them to engage with you on this issue and find out where your assessments could be adapted.

Encourage students to think critically about AI and its role in learning by asking questions such as: 

• What do you hope to learn at university? How do you expect to learn? 
• Who has heard of ChatGPT? Who has used it? 
• What are the expectations of how AI will be used in this subject? 
• What are the cognitive tasks that you should be able to complete without AI help, and when is it appropriate for you to use AI support? 
• What are some challenges about equity and AI?

Review your assessments

Below are some suggestions for minor adjustments that could be implemented in a reasonably short timeframe and should not require additional committee approval. Your discipline, direction by faculty, previously published material and equity issues may all impact on their suitability for your subject.

Provide clear guidelines on using AI in assessments

Be ready to answer questions about AI usage in your live classes. Students will need clear and consistent guidance on what they can and can’t do with AI in each subject. If AI is allowed as part of the assessment task, you should provide clear guidelines on when it is appropriate to use it (e.g. during draft development use only).

Also consider encouraging students to acknowledge when they use AI as part of referencing. If you are part of a teaching team, check with your Subject Coordinator to ensure your message is consistent across the cohort.

Because the text that ChatGPT generates is not retrievable or sharable, it falls into our catch-all ‘personal communication’ category, where you cite with an in-text only citation:

However, this is not an entirely satisfactory option, especially because the technology is so new, so both students and instructors are learning about this resource and how to ethically use it.

Consider, then, making the ChatGPT conversation retrievable by including the text as an appendix or as online supplemental material. If you do so, then readers may be referred to the appendix or the online supplemental material (where the ChatGPT response may be contextualized) when the ChatGPT conversation is cited.

It would be good practice to describe, in the narrative or a note, the prompt that generated the specific ChatGPT response. This too will help inform the understanding of the technology and its outputs.

Demonstrate ChatGPT in your classes

You could use an old question, or write one specially, prepare an annotated (‘marked’) version, and use that to show students the limitations of the AI, such as factual and referencing errors that would be obvious to a human marker. Compare this with the marking criteria for the task and explain why this answer would not do well without a lot more work.

You could extend this demonstration by outlining the amount of work it would take to ‘fact-check’ and correct the answer, then to produce accurate references. The subtext is that the AI doesn’t save time (at the moment) if you want a good quality answer, or possibly even a pass. Alternatively, you could challenge students to find all the errors in the output and correct them. Make sure you practice any such tasks yourself first to ensure it does demonstrate what you want it to.

Give an existing assessment a personal/current/unique context

If you currently have a written assessment task, consider setting a question that requires students to reflect on a personal, local or authentic (real life) aspect of their learning. You could frame a written assessment around unique in-person components of the subject, like guest presentations not available elsewhere, a field trip, laboratory task or a discussion or debate held in a live class during the current semester. Basing a task on very current content (at least post-2021) may provide a level of short-term protection. Be aware of the possibility that recent task materials pasted into AI may be integrated into the AI dataset. Incorporating visual prompts, unique datasets, multimedia resources or relating questions to specific texts may also provide some protection at the moment. For more information see reflective writing tasks.

Update assessment marking rubrics and marking guides for tutors

Consider the weighting of criteria within your marking guides and rubrics. Can you adjust these to place greater emphasis on criteria that assess learning beyond what AI tools are capable of producing? Identifying the key criteria and elevating their importance will require students to exercise evaluative judgement and undertake substantial revisions to attain a passing grade, even if they do use AI. Factual accuracy and correct referencing to appropriate scholarly sources (that exist!) could be given greater weighting as these are known weaknesses of AI (at present). Requiring all references to include active hyperlinks should highlight false references, both to students and markers.

Consider whether holistic rubrics would be more flexible than analytic rubrics in this evolving situation. If using analytic rubrics consider not allocating % or marks to particular criteria, or adjusting what you presently have. For more information see designing rubrics.

Sell the purpose of writing

Challenge students with a question like ‘Why learn to write in 2023?’, gather responses (perhaps using Mentimeter) and outline some reasons they might not have thought of:

• it forces you to organise your resources
• it enhances your subject matter expertise
• it deepens your understanding of your discipline
• it develops your disciplinary vocabulary
• it improves reading skills and vocabulary
• it translates into better speaking skills
• it demonstrates your ability to think clearly (BTW this makes you highly employable)
• it helps you learn about yourself (self-awareness is also a key career strength)

Remind students this is their opportunity to practice and learn in a supportive environment so they’ll be able to operate effectively in their chosen career. Outsourcing their learning will leave them vulnerable in the future workplace.

Build belonging to tap into values

The LX Adaptable Resources for Teaching with Technology provide practical strategies for Building Belonging into your first classes. This is essential when taking a holistic approach to students’ education and is proven to be a key indicator in long term motivation. You may also want to try to Support students with a values affirmation exercise that further reduces barriers to their success, particularly for students coming from under-represented backgrounds.

Ask for a reference list with links

If your assessment requires references, as students to include active links to each of their references. This might be as simple as a hyperlinked DOI like https://doi-org.ezproxy.lib.uts.edu.au/10.4324/9780429324178 or a permalink from the UTS Library.

This will assist markers, but also make students realise if the references provided by ChatGPT don’t exist or aren’t reliable. This is a key learning moment for students and part of their cultivating their own sense of academic integrity.