How Can I Support Students To Use Chatgpt To Support Their Learning

From a learning and teaching perspective, we should not assume that students will use any software that is not institutionally licensed because it may expose them to privacy and/or cybersecurity risks. However, knowing that some students want to use such tools, it is better if we teach them how to do so ethically and effectively.

1. Teach critical thinking: This is an academic skill that students are not born with, and teaching it will benefit their study of the use of chat beyond that, and it will enable them to make informed decisions about whether to use it (or other tools) and how to critique any data it provides.

2. Educate students to think critically about the use of chat, discuss the ethical issues surrounding the design and operation of the service, the personal, ethical and legal implications of using it so that students make informed choices about whether to use it, the potential impact on their learning and their educational outcomes in both positive and negative ways, and the professional implications.

3. Encourage learning how to learn. Understanding the process of learning is not an innate skill but needs to be learned. Students do not always understand that the activities or assessments they are asked to do are part of the learning process, so as not to have value to them. Be clear on how, and in what way, it will work to advance the learning of some participants so that they understand 1. How the application of AI might promote or impede that learning.

4. How it will be used for discussion. In the same way that many students (and staff) use Google Scholar, Wikipedia and other online sources, CHATGPT can also provide a starting point for research, provided students understand the limitations of its answers and how to find, use and reference additional sources.

5. How to clearly formulate questions. AI tools work on a Q&A model. They respond optimally to the question asked (unless they have insufficient data to provide a response). The process of revising the question requires students to think critically about the information they have discovered.

6. Students should also be able to critique the answers provided by the AI tool. An AI tool is not perfect. Data or poor reasoning in an answer will often be identified by the student as they would when critiquing any other source. Encouraging the student to critique the answers they get from the AI tool helps them develop a transferable skill that will only increase their employability.

7. Teach good academic practice. Support students to understand what good academic practice is, how submitting their own original work contributes to their learning and the value of this to them. Teach students how to appropriately reference, when and why it is needed. This is helpful in developing their understanding of these practices not just as a way to jump through hoops, but as important transferable skills.

8. Educate others of the dangers posed by chat. This tool works in a way where, at certain points, it detects lies and reports them as the truth. This is a danger for anyone without sufficient knowledge of what students can do to check against such lies, for example. Because of the way it was trained and the nature of its training bias, CHATGPT does not provide ethical, objective answers and in some cases answers that promote prejudice and discrimination, such as cultural, racial, gender.

Are there any risks to academic integrity and assessment security

As with any technology, there is the possibility that CHATGPT (and other similar tools) could be used to cheat. It is worth bearing in mind that not all students will cheat and even those who breach academic integrity may do so unintentionally (Burkes et al., 2020). There are many factors that can influence academic misconduct, ranging from individual factors such as less-developed academic literacy and skills, as well as what constitutes plagiarism/academic misconduct, to situational or contextual factors, including stress and mental health issues, other competing priorities (Morris, 2022a). Another significant threat to the security of an assessment is through staff’s use of CHATGPT (or its equivalent software).

Through this, you are both teaching the AI the questions and the correct answers. It then uses this information to provide insight to any student who subsequently asks it a similar question. This means that it will be more efficient to engage in good practice within assessment design, then to confirm, with colleagues and external examiners, if assessments are both effective and designed in a secure way.

What can I do to help protect the integrity and security of my assessments in an academic respect

A good starting point is talking to our students about academic integrity on an ongoing basis and supporting them to develop good pedagogical practice skills (Morris, 2022b). Assessments should be robustly designed and regularly updated. Many course teams already do this. It is easy enough to write an essay on a general topic for chat, for example, the contribution of technology to education; however, it is much less efficient to write in depth on specific topics, for example, the contribution of technology to your Heriot-Watt education between 2020-2022. Creating assessments that are more specific and/or more authentic requires students to relate their knowledge to personal or practical experience makes them more secure. After creating your assessment do not submit it to chat to “test” it, this will help you learn the answer, thus defeating your intended purpose.

Do I need to use AI detection tools

No, there are currently no accepted and effective tools that universities can rely on for reliable identification of fake AI generated content. As with all rapidly evolving learning technologies, this situation will be kept under close review for future developments. The risk to staff and student reputations due to misidentifications and false positives would be reputationally challenging.

Read Also: