This blog post is intended partly as a provocation, but one aimed at enhancing our understanding of ChatGPT (the most popular manifestation of generative AI) and hopefully inspiring critical engagement with the technology. I begin by reflecting on ChatGPT from a critical perspective, deploying the philosophical work of Harry Frankfurt On Bullshit (2005), to conclude that ChatGPT is the master bullshit artist of our time. I then propose the integration of three principles to promote the responsible and ethical use of ChatGPT: caution, trust, and transparency. I follow with an illustration of these principles at work, and conclude with a few remarks about where I think our future lies: inspiration and collaboration.

ChatGPT: The Bullshit Artist at Work

To illustrate the point that ChatGPT is a bullshit artist, let me tell you a quick story. In 2023, when confronted with the fact that students could make use of ChatGPT to produce their assignments, I realised that most students (and many colleagues) assumed that ChatGPT operates like a search engine, navigating the internet in search of information to answer their questions. But ChatGPT is not a search engine. ChatGPT is a language model that produces text based on statistical probabilities. To illustrate this to my students, I prompted ChatGPT to provide a profile of myself. Despite the fact that a detailed profile is publicly available in the University of Adelaide website, the profile generated by ChatGPT was riddled with errors and generalisations, albeit presented in very confident language.

This reminded me of the work of Harry Frankfurt On Bullshit (2005). I revisited his book and realised that ChatGPT was the perfect illustration of a bullshit artist at work. Frankfurt argues that bullshit is unavoidable whenever someone talks without knowing what they are talking about. This is exactly what ChatGPT does. ChatGPT generates text about anything and everything with no understanding of the content but, crucially, it does so with the confidence of an expert (this is often referred to as hallucination in the world of AI). In other words, ChatGPT is the master bullshit artist of our time.

Note that this does not mean that ChatGPT cannot produce accurate content. It can, and it often does. The point is that ChatGPT does not understand anything (it is a language model, not a knowledge model), and thus it cannot tell whether the content is accurate or not. This is why reliance on ChatGPT for learning and teaching is fraught with danger.

The Urgent Need for Normative Principles

The status of ChatGPT as the quintessential bullshit artist underscores the urgent need for normative principles to address the pedagogical challenges posed by its use in educational settings. Here, I am proposing the integration of three principles: caution, trust, and transparency. In essence, these principles amount to: (a) cautioning students about the use of ChatGPT (that is, making sure they understand how it works and the risks posed by its use); (b) developing a trusting environment between educators and learners (that is, having open and honest conversations about its use that inform the development of guidelines and expectations); and (c) encouraging and expecting students to be transparent when using ChatGPT in their assignments.

The Principles at Work: ChatGPT Appendix

In my teaching practice, these principles translate into a requirement for students to include a ChatGPT Appendix (which I have extended to include other AI tools) if they have used generative AI in the production of their essays. I reinforce the communication of this requirement using the advice provided by Australian Customs and Biosecurity to people who arrive in Australia: "Don’t Be Sorry, Just Declare It.” Students are reminded of this basic expectation of transparency in the assignment submission portal:

Don't forget to include a ChatGPT / AI Appendix at the end of the essay, after the Reference List, if you have used ChatGPT / AI in the production of the essay. The absence of this Appendix is equivalent to stating: "I did NOT use ChatGPT / AI". If this statement turns out to be false, this would constitute a breach of Academic Integrity. Remember the slogan: "Don't Be Sorry, Just Declare It".

The available evidence, drawn from the implementation of this approach in a course with 113 enrolments, indicates that students respond well to this pedagogical initiative. Significantly, the evidence also suggests that this approach can minimise the cases of academic misconduct. In this course, most students chose not to use ChatGPT and those who used it did so in limited and acceptable ways (e.g. assist with expression, idea generation, improve understanding). There were only three essays that displayed problematic uses and required follow up conversations.

The Future of Us: Inspiration and Collaboration

I hope this blog post can inspire educators to examine their own views and practices regarding ChatGPT and develop an interest in the pedagogical uses of AI in general. I am open to explore collaboration opportunities with colleagues who are interested in either of these two aspects of this project, namely: understanding the implications of using generative AI tools in pedagogical contexts; and/or exploring and developing principles for the responsible and ethical use of generative AI in learning and teaching.

The way in which we respond to the profound societal challenges posed by the rapid development of generative AI will shape our future, including the future of higher education. Addressing these challenges calls for inspiration and collaboration among educators, technologists, and policymakers, to create frameworks that promote the responsible use of this emerging technology, while harnessing its pedagogical potential.

Note: This blog post draws on content from the award-winning digital and onsite posters presented at the HERDSA Annual Conference held in Adelaide in July 2024.

Source of banner image: https://hawar.no/2023/04/what-is-chatgpt-how-do-you-use-it-how-do-you-be...