Artificial Intelligence

AI in my Learning

I am a great believer in maintaining familiarity with the learning experience to understand students' perspectives and to inform my practice. Thus, whilst learning to use Excel, I used generative AI as a 'just-in-time study buddy' to clarify my understanding of Excel logic, debug formulas, and provide feedback on my solutions. As my university permitted the use of GenAI, I also utilised it to create exam-style practice questions, allowing me to engage in distributed practice testing. I quickly learned that GenAI is simultaneously masterful and inept, as it produces 'hallucinations' and may generate convoluted formulas. Much of its usefulness also depends on the sources it draws from and the type and quality of the prompt. I also learned that ChatGPT is optimised for helpfulness, cooperation and safety over debate and accuracy, as demonstrated when I told it that it was incorrect when it was not. I found that this often results in inaccurate, biased, "locally shaped" and misleading information, as shown in the images below.

This issue was also highlighted during a CPD event I attended in which attendees were prompted to compare Claude.ai and ChatGPT answers to a question on the topic of race. It was very interesting to observe the bias and simplicity in ChatGPT's answer resulting from its prioritisation of cooperation in comparison to the answer provided by Claude.ai, which was more insightful, nuanced and balanced. It was also noticeable how the answers reflected certain cultural and societal biases that may not be applicable in other societies. I found this difference to be quite interesting, and it prompted me to pay closer attention to the ethics of AI and raise this issue among my students in more explorative and experiential ways.

AI in my Teaching