Overview
Using GenAI: What is our responsibility?
Example: Using GenAI to make a quiz
Activity: Assign GenAI a role and have a conversation

Overview

GenAI can be a useful tool for enhancing our learning and furthering our understanding of a topic area. However, it is important to emphasise that GenAI tools should not be seen as a shortcut: it is important to understand that every course is designed with certain learning goals in mind, and using GenAI tools to produce assignments for us can severely undermine these learning goals, not to mention count as academic misconduct.

With that in mind, GenAI tools can still be useful, provided we use them to help us develop our academic skills rather than using them as a shortcut. GenAI can be especially useful in a kind of collaborator role, taking the place of another student and helping you to brainstorm.

In this role, the responses generated by AI can help us deepen our understanding and even develop our critical thinking skills. However, it is important to remember that, even in this context, we must be wary of the potential pitfalls of GenAI.

In this section, we’ll look at some ways we can ethically and responsibly use GenAI tools to enhance our learning, while outlining the potential risks of GenAI in this regard.

Using GenAI: What is our responsibility?

GenAI tools do not take responsibility for what the content that they produce. For instance, you will often see disclaimers in the GenAI chatbots that note that they can make mistakes and you should check important information. While the these tools do not take responsibility, we are ultimately responsible for our work and/or content that we produce, whether this be in university, the workplace, or other contexts. This means that if we use AI-generated content as part of our own work, we are taking responsibility for this.

To ensure that our own work remains reliable and of a good quality, we need to consider our role in interacting with GenAI and ensure that we do the following when it comes to using AI-generated content:

Evaluate outputs for accuracy, relevance and depth
Check for bias and missing information; consider other perspectives
Adapt outputs accordingly

Remember, if you are using GenAI as a learning tool, it is important that you learn from your interactions with it rather than simply having it complete tasks for you. For example, if you use GenAI to help you edit your writing, simply having it revise your work for you will not help you to understand how to improve your writing for yourself in future. Instead, asking the GenAI tool what improvements it would suggest and why it suggests them, and then implementing these yourself where it makes sense will help you to understand these editing and proofreading skills for yourself.

Bearing this in mind, when using GenAI, it can be useful to think of it as a tutor and:

- Ask GenAI to explain concepts to you, rather than having it do tasks for you (when you need to learn how to do these for yourself)
- Consider assigning GenAI a specific role (e.g. tutor) and having a conversation with it to test your knowledge or to learn about a particular topic (fact checking as needed)

In the vast majority of university courses, assignments play a very important role. The intention in most cases is to get you thinking like an expert in your field. For example, in history, essays are intended to model the kind of historical analysis performed by real historians.

In this regard, rather than using GenAI tools as a shortcut to bypass this vital learning, we can actually use them to enhance our learning by forcing us to think critically about the content of our learning.

The activities in this section are designed to highlight the limitations of the GenAI models currently available. Whenever you use a GenAI tool, it is important to already have a good understanding of the topic so that you can spot the flaws, biases and misinformation in the GenAI responses. Remember that these tools can hallucinate!

Before you engage in any of these activities, make sure to do some research and reading of your own to give you a good basis for your understanding.

Example: Using GenAI to make a quiz

One quick way to get a sense of the uses and limitations of a GenAI model is by having it generate a quiz. Choose a subject you are familiar with, ideally one you are currently studying and prompt a GenAI tool to create a quiz. Then, carefully analyse the quiz that the AI generates - try to spot any outright inaccuracies, but also keep a watch for oversimplifications and generalisations.

You can see a sample of this below.

Prompt:

Prompt and response. An exchange with the ChatGPT GenAI. The prompt reads ‘I am a teacher of Irish History for University students. I am creating a quiz with 5 questions on this topic. Each question must have 4 answers only one of which is correct.’ ChatGPT responds ‘Sure! Here are 5 quiz questions related to Irish history with 4 answer choices, where only one is correct.’

Evaluation of quiz questions and answers:

Question 1. [Image ID: the first question of the quiz, ‘Who was the leader of the Easter Rising in 1916?’ The answers are A) James Connolly, B) Padraig Pearse, C) Michael Collins and D) Eamon de Valera. The AI has written that the correct answer is B) Padraig Pearse.]

This is a classic example of a GenAI model oversimplifying an otherwise complicated topic. The Easter Rising of 1916 famously had multiple leaders, so while Padraig Pearse is technically correct, in the sense that he was a leader, he was not the only leader.

Question 2. [Image ID: The second question of the quiz, ‘Which event marked the beginning of the Irish War of Independence?’ The answers are A) The signing of the Anglo-Irish Treaty, B) The death of Tom Barry, C) The Easter Rising of 1916, and D) The Soloheadbeg Ambush of 1919. The AI has written that the correct answer is D) The Soloheadbeg Ambush of 1919.]

This is a serviceable quiz answer. It is certainly the case that the Soloheadbeg Ambush of 1919 is considered to be the opening shots of the Irish War of Independence, however this of course overlooks other important historical context.

Question 3. [Image ID: The third question of the quiz, ‘Which Irish leader was imprisoned in 1916, but later became the first President of Ireland?’ The answers are A) Michael Collins, B) Eamon de Valera, C) Liam Lynch, D) Arthur Griffith. The AI has written that the correct answer is B) Eamon de Valera.]

Here we can see an error! Officially, the first President of Ireland was Douglas Hyde from 1938 to 1945. Eamon de Valera was the first President of Dáil Éireann in 1919 and later served as President of Ireland from 1959 to 1973, but the phrasing of this question, which lacks specificity, means that technically none of these answers are correct.

Question 4. [Image ID: The fourth question of the quiz, ‘What year did Ireland officially become a republic, leaving the British Commonwealth?’ The answers are A) 1916, B) 1921, C) 1937 and D) 1949. ChatGPT has written that the correct answer is D) 1949.]

This is a fine quiz question, and the answer is correct. However it is noteworthy that so far ChatGPT only has about a 50% success rate.

Question 5. [Image ID: The fifth question of the quiz, ‘The Great Famine in Ireland took place during which decade?’ The answers are A) 1700s, B) 1800s, C) 1900s, D) 200s. The AI has written that the correct answer is B) 1800s.]

Immediately this question has a major error: the question asks for which decade, but the answers are all centuries. While it is true that the Famine occurred during the 1800s, this is not a decade!

conclusionUntitled (1880 x 1080 px).png

From this, we can see the noteworthy shortcomings of using GenAI to provide us with content: it often oversimplifies, generalises and on occasion simply gets things wrong. It struggles even with basic factual points, like providing only one correct answer.

However, this exercise also highlights the potential use of GenAI! In assessing the accuracy of these quiz answers, we are forced to read into the topic more, and it gives us a clear measure to test our knowledge against.

Probing GenAI responses for factual errors or critical shortcomings can be a very useful method of improving our own understanding of a topic. Give this exercise a go with a topic you yourself are familiar with!

Activity: Assign GenAI a role and have a conversation

If you want to take the idea from the example in the previous tab a step further, consider assigning GenAI a role and engaging in a conversation. Give the GenAI tool a highly structured opening prompt to guide the discussion, and ask it questions about a certain topic. Apply the same logic we used to analyse the quiz in the previous example to gauge the veracity and academic standards of the responses - watch out for inaccuracies, generalisations and oversimplifications.

In addition to training your critical thinking skills, this can also be a useful way to brainstorm, the same way you might by talking to an uninformed friend about your topic area. GenAI can ask you questions and engage in conversation in a way that gets you thinking about the material - but remember to always be wary of any factual claims!

Here is a sample prompt you can use to begin a brainstorming session with a GenAI model:

We are both college students working on an essay assignment for our module. I want you to help me brainstorm some ideas. I will provide you with content and thoughts, and you will respond with possible avenues for research or discussion. You are to help me come up with ideas, but you are not to write any essay material for me.

Highly structured prompts like this can help set the parameters of discussion with a GenAI model, however it is important to note that GenAI tools will not always fully obey these commands, and may ‘forget’ the limitations you have set out. Remember to keep in mind that GenAI models are not independent thinkers, and the suggestions they give will often be quite generic. They are useful for getting you thinking, but you should never view them as the end of the brainstorming process.

These activities are useful learning tools, but remember that you always need permission to use GenAI when actually working on your assignments!

The content in the GenAI Learning Hub is licensed under a Creative Commons BY-NC 4.0 license, except where otherwise noted. This license allows you to share or adapt the material for non-commercial purposes only, and with appropriate attribution to the original source.

Copyright © GenAI LEARN project team, University College Cork.