Student reflections on how GenAI supports learning in a Graduate Hydrology course

Course: EOSC 533, Advanced Groundwater Hydrology

Instructor: Dr. Roger Beckie, Professor, Department of Earth, Ocean and Atmospheric Sciences, Faculty of Science


In EOSC 533, Advanced Groundwater Hydrology, Dr. Roger Beckie encouraged his 36 students to reflect on their experience with GenAI, specifically how students use AI to enhance their learning. He shared a blog post with his students by Sandrine Desforges, Research Associate at the Higher Education Strategy Associates (HESA) that highlights recent developments in the use of artificial intelligence (AI) in post-secondary education, featuring new resources such as:

  • The Levels of AI Assistance framework (student guide) and
  • The revised AI Assessment Scale (on a scale that spans from No AI to AI Exploration)

Reflection questions

Here are the questions students were asked to reflect upon and share:

Experiences

  • Which GenAI tools have you used in your studies and why: ChatGPT, Microsoft Copilot, Gemini, Bard, DALL-E, etc.?
  • Describe a specific use case where GenAI affected your learning and learning workflow.
  • What prompts did you use? Was it effective?

Workflow

  • Do you use GenAI in your regular workflows? If so, describe them.
  • What strategies or prompts do you find most effective?
  • Can you recommend any resources to learn more about using GenAI to enhance learning?

Challenges

  • What challenges have you found using GenAI tools?
  • How did you overcome these challenges?

Key takeaways from student responses

Dr. Beckie reminded students that they could use Microsoft Copilot with their CWL credentials. Here are the key points that emerged from student responses, which reflect how GenAI tools could fit into their learning experience.

Experiences

  • Use in studies and work: Students employed ChatGPT and Microsoft Copilot for writing assistance, brainstorming, grammar refinement, and summarization of complex topics (e.g., rock mechanics, programming). Generative AI has proven useful for code development, troubleshooting, and learning complex concepts, though familiarity with the topic enhances success.
  • Learning efficiency: AI tools facilitate understanding by breaking down concepts, offering structure to reports, essays, and coding. Students appreciate AI’s ability to simplify complicated topics, generate study questions, and suggest structures for papers and projects.
  • Limitations with complex topics: While AI tools can summarize or offer creative suggestions, they tend to struggle with industry-specific, technical, or scientific accuracy.

Workflow

  • Enhanced productivity: AI helps automate repetitive tasks such as generating Excel formulas, drafting emails, and providing creative content (e.g., titles for meetings). In coding workflows, users utilize AI to generate initial script versions, which they refine through testing.
  • Iterative use: AI is mostly used as a starting point for content creation, where users improve and customize generated results. Prompts like “explain this code” or “suggest a title” lead to multiple rounds of refinement for better results.
  • Complementing workflows: AI assists with brainstorming, ideation, proofreading, and providing visual aids (e.g., video generation). However, users maintain the habit of reviewing, testing, and adjusting AI outputs before finalizing them.

Challenges

  • Accuracy and reliability issues: Students face challenges with AI’s frequent “hallucinations” (fabricating sources, papers, or data). The need to double-check AI-generated content is critical, especially in fields requiring precision (e.g., engineering).
  • Complex prompts lead to errors: When tasks or topics become more technical or nuanced, AI often produces inaccurate, incomplete, or oversimplified responses, necessitating manual intervention.
  • Confidentiality concerns: Many students report challenges integrating AI into workflows involving sensitive or proprietary information due to security restrictions imposed by employers, limiting the use of AI in professional settings.

Dr. Beckie’s reflections on GenAI in education

Based on those comments, Dr. Beckie has not made any conclusions yet about the teaching approach to take, but he has some ideas. First, he likes the approach of learning by mimicry, and the idea of providing examples to students how AI can help them learn. What prompts are effective?

He thinks it could really help leverage and scale instructor feedback, which we know is critical for learning. He was also thinking that something like NotebookLM could be useful, where students could link course reading(s) in the notebook, and then ask the AI questions. The AI would use the sources provided but also the full power of Google Gemini to bring in other literature. This is something he is exploring for this course.