The Complex Relationship between AI and Education
The Complex Relationship between AI and Education
Alongside her colleagues across Boston University, Naomi Caselli is tackling some of the thorniest questions about the relationship between artificial intelligence and education. Caselli, an assistant professor of Deaf education at BU Wheelock, is co-director of the AI & Education Initiative, a research initiative led by a coalition of faculty from across the University, including BU Wheelock, the College of Arts & Sciences, the College of Engineering, and Metropolitan College.
Housed at BU’s Rafik B. Hariri Institute for Computing and Computational Science & Engineering, the AI & Education Initiative includes researchers like Caselli who are already working on studies using AI to evaluate career-readiness policies, develop sign language technology, assess reading, and study writing education. Together, their goal is to understand the complex relationships between AI and education—and develop strategies for making use of these new technologies in the classroom.
Shifting cultural norms
The rise of interactive AI programs like ChatGPT has underscored the importance of the work of the AI & Education Initiative. Though much of the attention surrounding ChatGPT and similar technologies has centered on risks, like the potential for cheating on written assignments, Caselli advocates a more nuanced approach to understanding the role of natural-language AI in education.
This approach involves anticipating a shift in cultural norms—and adapting pedagogy to match. When everyone has access to augmented writing, Caselli thinks debates about cheating, plagiarism, and academic honesty will change.
“Figuring out how to use text generation responsibly is fraught,” says Caselli. “As with any new invention, it takes a while to come up with cultural guardrails.”
Recognizing new norms means teaching students new ways to approach media literacy, too. ChatGPT can be a useful tool, but its inability to judge the quality of data can also lead it to produce answers that sound convincing, but may ultimately be false. In addition, since natural-language AI is based on large data sets, common stereotypes and misconceptions—or outright falsehoods—may emerge from the program’s responses.
“How do you fact-check ChatGPT when it’s wrong?” Caselli asks. “And how do we tweak it so that it’s right?”
The promise of AI for learning
For Caselli, the possibilities that ChatGPT and other natural-language-processing AI present are wide and varied. For example, teachers can use an AI to become more effective in the classroom. They can also use natural-language technology to help with complex paperwork, putting the finishing touches on educational plans for students with disabilities, lesson plans, writing prompts, and test questions.
Students, too, can benefit from augmented-learning tools like ChatGPT. Just as social science researchers use statistical software packages instead of pencil-and-paper calculations, students can use AI to support them in completing formulaic written assignments, such as book reports.
“It will become normal to use AI. It’ll be understood that most of our writing is augmented that way, in the same way that we use spell-check, calculators, and statistical software.”
If they don’t understand a lesson, they can ask the AI to reframe the content using different words, or change the writing style to something more understandable.
Natural-language AI can also help students take notes more effectively—and they can be especially helpful for students who may need accommodations in the classroom, including some of Caselli’s own students.
“Deaf students can’t easily divide their attention to look at a teacher or interpreter signing, write notes, and look at a PowerPoint at the same time,” she says. “Students could audio record the lecture or use AI to automatically transcribe the audio, and have ChatGPT summarize pages of transcripts into usable, brief notes.”
Ultimately, despite the complicated issues associated with ChatGPT and similar AI tools, Caselli is optimistic about the long-term educational benefits.
“It will become normal to use AI,” she predicts. “It’ll be understood that most of our writing is augmented that way, in the same way that we use spell-check, calculators, and statistical software.”