Artificial intelligence is rapidly reshaping the classroom experience at the University of Toronto, prompting professors to overhaul traditional teaching and evaluation methods as student reliance on tools like ChatGPT and Google’s Gemini accelerates. The shift became clear to U of T Scarborough associate professor André Cire two years ago, when he noticed a widening gap between high marks on take-home assignments and declining performance on in-person exams. The explanation, he said, was straightforward: students had begun integrating large language models into their coursework at unprecedented rates.
Polls from KPMG Canada reveal that in 2024, about one in six Canadian students over age 18 used AI tools for schoolwork, a sharp 13 per cent rise from the previous year. Faculty members who spoke with TorontoToday described the trend as both helpful and concerning. Some students use AI as a study companion that provides instant feedback and explanations. Others use it as a shortcut to complete assignments without mastering the underlying material, blurring the line between assistance and academic misconduct.
Professors say the signs of AI use are becoming increasingly recognizable. Cire said student writing has grown more elaborate, with unnecessary flourishes that do not match their in-class abilities. Coding assignments once closely mirrored class examples; now they arrive longer and more polished, reflecting the influence of generative models. Longtime computer science professor Karen Reid said it is almost impossible to reliably detect AI-generated work using current tools, though some students have made obvious mistakes — including submitting assignments containing leftover ChatGPT phrasing. Under U of T’s academic code, using AI on assignments where it has been prohibited counts as an unauthorized aid, the most frequently cited form of cheating since 2020.
Beyond academic integrity concerns, AI is also changing how students interact with one another and with faculty. Professors report declining attendance at office hours and reduced participation on online course forums. Reid noted that AI offers round-the-clock answers that a human instructor cannot match, yet she worries that students may be developing an inflated sense of understanding based on overly agreeable model responses. Assistant professor Rahul Krishnan echoed the concern, cautioning that easy, instant explanations risk eroding the essential skill of learning how to learn.
In response, U of T faculty are rethinking both course design and evaluation. Some, like cinema studies professor Bliss Cua Lim, are shifting toward assignments that rely on personal engagement and cannot be easily outsourced to AI, such as interviewing family members about past film-going experiences. Others are adopting multi-step assessments that force students to demonstrate genuine comprehension. Cua Lim now requires students to annotate readings before writing reflection papers, resulting in deeper engagement with both text and interpretation.
In fields like data science, experiential learning has taken on new importance. Cire now brings business executives into the classroom and asks students to solve real-world problems, making it far more difficult to rely on AI tools alone. He has also rebalanced grading by incorporating in-person quizzes and oral assessments worth roughly 30 per cent of the overall mark. Since adopting these changes, he said student engagement has improved and grades have stabilized.
Professors agree that as AI capabilities evolve, teaching will have to evolve with them. The challenge, they say, is not eliminating AI from academic life but finding ways to ensure students continue to develop critical thinking and independent learning skills. As Cire put it, the pace of change means the work of adaptation will be ongoing: “Maybe next summer, if you ask me again, I’ll be doing something completely different.”
