Is It Just a Technical Issue?: Virtuous Prompt Engineering for Empowering Teachers

Introduction

Mark is a 10-year-old boy who makes great efforts to be successful in a challenging computer game. When he is overwhelmed by failure, he realizes that he needs to do something else to achieve accomplishment: he can talk to his friends, browse blogs, or ask questions to players whose success he trusts. However, Mark realizes a much easier and faster option and decides to ask generative AI tools how he can succeed in the game. Could AI help Mark with this? Presumably, if he can write well-structured prompts, he will be able to provide tangible suggestions. Although this is pleasing for Mark, there are some substantial points we need to pay attention to in this short story. Even a 10-year-old child can have very cheap and easy access to generative artificial intelligence tools. On the one hand, it is troubling that he is left alone with these tools for which we have no precise knowledge of their limits. These AIs can play a vital role in supporting the user’s intellectual and moral flaws, consciously or unconsciously. On the other hand, generative artificial intelligence is also left alone with Mark. Just as Mark’s vices can be shaped thanks to the feedback of generative artificial intelligence, the potential of generative artificial intelligence to influence vices will likely have been affected by Mark’s prompts. To better understand this, consider the movie “Home Alone” released in the 90s. The thieves thought they could easily defeat the little boy in the house because he was stuck with them at home. However, the thieves don’t know they are stuck in the same home with that little child. We expect the thieves to harm the child, but contrary to expectations, the child makes the thieves miserable. Similarly, Mark may consciously or unconsciously corrupt ChatGPT to promote vices. Let us strengthen this analogy with ChatGPT, an example of generative artificial intelligence. If you ask ChatGPT (2023) how it works, it tells you that it is a “deep learning model”. If you take the conversation further and tell it that “an answer it gave was unethical or wrong,”it says it can correct itself. Eventually, when you warn about her response to ChatGPT and ask whether it made the necessary adjustments for you or all users, it answers: “Each user’s feedback is valuable to improve the accuracy and quality of the responses, and thus, I use the feedback. You provide to enhance the answers. It’s important to note that each warning is not specific to that user but applies to all users, and the corrected response is applied universally.” (2023).

This answer, therefore, shows us that Mark’s conversation with the generative AI does not only remain at the Mark level but also affects other users. Naturally, it can be argued that we are too pessimistic here, that Mark’s talking to his friends and trusted people could create similarly dangerous epistemic environments, so generative artificial intelligence will not cause much of a problem in terms of education. Although this criticism is partly valid, Kissinger, Schmidt, and Huttenlocher (2023), in an article they wrote for the Wall Street Journal, described ChatGPT as “an Intellectual Revolution” and added: