Several members of the ASA Section on Text Analysis are interested in the performance and impact of chatGPT. The section officers invite members to share their experiments with the chatbot and their ideas about its influence via this messaging platform.
David Banks used the prompt "Describe moral philosophy, including consquentialism, deontology, and virtue ethics, in terms of Marvel superheroes." He obtained the following:
"Moral philosophy deals with ethical questions and principles, such as what is right and wrong, good and evil, and justice.
Consequentialism evaluates actions based on their consequences, where the rightness of an action is determined by the overall outcomes it produces. For example, in Marvel, Iron Man can be seen as a consequentialist as he makes decisions based on the results they will produce, even if they may not align with strict moral codes.
Deontology, on the other hand, holds that there are certain moral duties and rules that must be followed regardless of the consequences. Captain America is a good example of a deontologist, as he follows a strict code of ethics and sticks to his principles, even if it goes against his personal interests or the greater good.
Virtue ethics focuses on the character and habits of the moral agent, rather than rules or consequences, as the key element of ethical thinking. This approach values moral character traits like bravery, honesty, and compassion. For instance, Spider-Man can be considered a virtue ethicist as he strives to embody these traits and use his powers for good, regardless of the situation he is faced with."
On the other hand, although chatGPT is able to correctly define Petrarchian, Spencerian and Shakespearian sonnets, when given the prompt to produce a sonnet on the significance of John Donne's poetry in each of these forms, it produced four quatrains, not sonnets, all with the same rhyme structure.
I conjectured that chatGPT might not be able to describe the difference between finger tight and tool tight. Every mechanic knows this intuitively, but I felt this might be difficult to capture from internet text. So I gave it the prompt "Write a poem that defines the difference between finger tight and tool tight" and it generated
"Finger tight, a gentle hold,
Just enough to keep it bold.
With ease it can be moved around,
A light touch, not too profound.
Tool tight, a different story,
For stability, it's all the glory.
Tightened down with all one's might,
It won't budge, withstanding any fight.
Finger tight's for quick adjustments,
Tool tight's for heavy loads and enhancements.
One's for ease, the other for strength,
Choosing which, depends on the length.
So when you need a secure bind,
Rely on your tool, be tool-tight inclined.
And when a gentle touch is best,
Remember finger tight will do the rest."
Finally, there has been much chatter about the threats chatGPT may pose, especially in education with the possibility of easy-access automatic essays. This is a concern, of course. Some people claim to have software that can detect essays generated by (today's) chatGPT---see https://www.npr.org/2023/01/09/1147549845/gptzero-ai-chatgpt-edward-tian-plagiarism. However, I am interested in the possibility that next-generation chatGPT might be able to provide personalized tutoring to K-12 students, tailored to each student's interests and academic preparation, anywhere in the world. If that is possible, it could have very positive results.
------------------------------
David Banks
Duke University
------------------------------