The rapid adoption of AI into the classroom has even extended to a flagship computer science course at Harvard University, which now uses ChatGPT as a way to free up teaching assistants to spend more quality time with their students.
It’s one of the newest applications of artificial intelligence, a technology that has caused both excitement and alarm among educators since the release of OpenAI’s ChatGPT in late 2022. The ripple effect of AI is affecting everything from entrance exams to how teachers create assignments.
Harvard’s Computer Science 50: Introduction to Computer Science introduced AI as a tool in its summer program about two weeks ago. This popular course has about 70 students this summer, and he expects to have over 600 in the fall.
“We’re pivoting heavily on AI and what we see as potential benefits,” said David Mullan, a professor at Gordon McKay and computer science practice. He said computer science classes really need more guidance for students.
“Even with the generous resources that we have today, it’s not enough,” Maran said. “The hope is to support students as much as possible through software and reassign the most useful resources — people — to help those who need it most,” he said. Instead of reducing the number of teachers, increase the number of teachers. ”
Building on ChatGPT, Harvard uses this technology to help computer science students understand highlighted lines of code and give advice on why and how to improve the style of their code. It is also used to answer frequently asked questions.
Maran said the technology is not intended to replace TAs or professors, but to help them make better use of their time.
Lynn Parker, director of the AI Tennessee Initiative, says many institutions are facing a shortage of TAs, especially in computer science, so leveraging AI in this way could be a positive change. I think.
“I’m not worried about losing existing jobs,” said Parker, who is also vice president of the University of Tennessee, Knoxville.
“It helps extend this kind of assistant to more students,” she said. “And having all students easily access personalized help in their courses is a great goal.”
Harvard University is the latest in a growing list of universities to adopt AI. Admissions offices, including Georgia Tech, are “experimenting” the technology. Other educational institutions are also launching his AI teacher workshops. And some universities, such as Purdue, Emory, and Albany, employ dozens of faculty members who specialize exclusively in artificial intelligence.
At Harvard, Maran said its current use in introductory computer science is just the beginning. He also described four additional applications of his that he hopes to deploy in the near future. The AI explains errors his messages, helps students find bugs in their code, evaluates program design, and evaluates students’ comprehension through AI conversations.
These future features will focus on pointing students in the right direction by asking rhetorical questions, like a TA, rather than telling students bluntly what’s wrong with their code.
The Harvard class could set the standard for wider adoption of AI, especially among the tech-savvy computer population, Parker said.
Having easy access to personalized help within a course for every student is a great goal. ”
— Lynn Parker
But for some faculty, it may take a little longer to convince them, says Fiona Hollands, founder of the research evaluation group Ed Researcher.
“Some faculty are saying, ‘We’ve been doing it this way for 20 years, why are we changing it?'” said Hollands, a former research consultant with the Center for Technology and School Transformation at Columbia University Teachers College.
Hollands said it was difficult to get teachers to do things online before the COVID-19 pandemic created “a compelling situation.” “Higher education institutions don’t tend to look at ‘how can we be more efficient?'” she said.
Maran pointed to the potential of AI to assess students’ code designs, but this is largely subjective and time consuming for TAs. TAs spend hours grading designs, but on average, students read TAs’ comments in less than 14 seconds, he said.
According to Maran, time was never a good use for TAs and students. “While we do not intend to completely eliminate humans from the process, their impact on the student experience is amplified.”
Easy access to AI not only frees up TA, but also allows students to troubleshoot faster instead of becoming increasingly frustrated while waiting for TA to become available. purpose.
“It’s possible to use this to break deadlocks, and this is what you do with human assistants, but often they aren’t available,” Parker said. “So often students get stuck and say, ‘I’m giving up’ and quit. [This technology] may help maintain. ”
Maran emphasized that AI is not omnipotent and continues to evolve, and said students are reminded of this when AI answers questions. Literal footnotes attached to each response warn students not to take responses at face value unless approved by the TA.
The AI application is currently only used in computer science courses, but Maran hopes to roll it out to other courses this fall, putting it outside the walls of typical STEM.
“My hope is to collaborate outside of computer science and STEM, and specifically look at the humanities as another proof of concept,” Maran said. He said he wanted to show that “this is not just STEM he’s in, he’s not just in STEM.” It has an overall positive impact. ”
