An attempt to incorporate ``how to use and how to evaluate'' into student education instead of prohibiting AI

Image generation AI such as Stable Diffusion and Midjourney, interactive AI such as ChatGPT and Google Bard have appeared, and AI can generate pictures and sentences that could only be created by humans with human-level accuracy. became. Of course, there are also voices of doubt about its accuracy and reliability, and there is also a report that `

`17% of students use ChatGPT for exams and assignments, '' and there is an opinion that the use of AI in educational settings should be prohibited. is also chanted. Under such circumstances, attempts are being made to teach students how to use and evaluate automatically generated AI at universities and high schools.

My class required AI. Here's what I've learned so far.

At This School, Computer Science Class Now Includes Critiquing Chatbots - The New York Times

Professor Ethan Morrick of the Wharton School of the University of Pennsylvania challenged his students to use AI in their undergraduate and graduate courses. In one class, he instructed to use AI for creating ideas, writing sentences, creating applications, generating images, etc. Also, in another task, we required or permitted the use of AI to clear the task. In another class, he introduced AI and instructed him to think about how to use it.

According to Professor Morrick, other lecturers and professors have evaluated that 'ChatGPT itself can write quite natural sentences, but there are many terrible writing tasks using AI.' When I actually let students use AI, it seems that most of the cases were not able to master AI at first.

For example, Professor Morrick gave students the task of ``composing a five-paragraph essay on a given topic''. Students were forced to use AI, and they let AI do the composition as it was, or made an illustration suitable for an essay. However, when I actually read the essay, most of it was mediocre, and my grades were C minus at most.

by Mike MacKenzie

Professor Morrick seems to have instructed to enter multiple command sentences (prompts) to be entered into ChatGPT when issuing assignments. So, when checking the prompts entered by the students, roughly ``1: leave everything to AI'', ``2: the user decides what the AI will write and also indicates the range of knowledge that he knows'', and ``3: AI After checking the content written by , I will proofread the sentences and rewrite them in detail.”

The essay written by the first approach will inevitably be mediocre content. The second approach was better than the essays written with the first approach, but the results were more variable. And it seems that the best content was the one output by the third approach.

After having the students submit an essay, Professor Morrick taught again how to instruct AI based on the third approach, and finally the completion of the essay was dramatically improved.

Many educators are concerned that AI is unreliable. It seems that the content was carefully fact-checked. However, since the accuracy of the fact check itself is low, the possibility of being deceived by ChatGPT should be considered more.

Also, at a public high school in New York, a computer science teacher teaches `` critical computing '' classes. In this class, we aim to understand AI based on critical thinking , not just computer programming.

In this class, for example, we discuss how a facial recognition AI developed by a tech company team made up mostly of Caucasian and Asian men can have difficulty recognizing black and Latino faces. He also said that he is discussing what to do to solve the bias seen in AI.

In addition, the instructor of this class used ChatGPT to create a 30-minute lesson plan and proceeded with the lesson accordingly. Students discuss the usefulness and problems of the ChatGPT class by comparing it with the class plan actually created by the instructor.

Professor Morrick said, 'Even if I don't embrace AI, it's clear that AI is everywhere around students. They help me complete class assignments even before I teach them how to use it.' We used AI as a tool,” he commented, saying that without training, everyone would use AI incorrectly, so instead of focusing on whether or not to use AI, we should learn by focusing on how to use AI. It is expected that the effect will improve.

in Note,   Software, Posted by log1i_yk