Getting Started with AI

At its core, Gemini functions as a conversational collaborator. You interact with it by asking questions, making requests, or uploading files — and asking Gemini to process, analyze or transform that information.

The requests and instructions you give to Gemini are known as "prompts," and the clearer your prompts are, the better the outcomes. To keep your prompts focused and effective, follow this basic structure:

• Role: Begin by stating the role that Gemini should take. For example: “You are an expert guide on active learning.”
• Task: Describe the task or goal you want Gemini to pursue, using action verbs. For example: “Your goal is to help me brainstorm active learning activities for my course."
• Context: Provide Gemini with as much contextual information as you can. For example: "I am teaching advanced undergraduate microbiology. The course is in-person, I have 125 students and 2 TAs for support."
• Format: Tell Gemini what kind of interaction and what final outputs you want from it. For example: "Ask me questions about my learning outcomes until you have enough information to suggest some aligned active learning activities. I also want you to explain why you think your suggestions are a good match."
• Constraints: For some tasks, you may want to set limits or requirements for Gemini. For example: "Give me no more than 3 suggestions, only make suggestions that will work in a lecture theatre with immovable seats, and limit your explanations to 100 words."

Test your prompt in Gemini. If you don't get the output you expected, refine your prompt and try again. Prompting is a skill that improves with practice — the more specific and intentional you are, the more clear, cohesive, and useful the results will be.

"Gems" take the general capabilities of Gemini and lock in your specific instructions. For instance, you might create and save a gem that helps students plan their approach to a final project — or a gem that plays the role of a patient, enabling students to practice their clinical assessment skills — or a gem that helps students clarify a draft without rewriting it for them.

Building a gem is much the same as "prompting" Gemini (see above), except that you can save your pre-prompted 'flavour' of Gemini and share it with others.

If you'd like to try building a gem, follow the instructions in this short guide and 3-minute video. You can also register for an upcoming AI Buildspace, where experienced colleagues will coach you along in a live workshop setting.

The datasets used to train GenAI models contain biases and harmful stereotypes, in part because the vast majority of the training corpora used for LLMs are heavily English-centric and Western-oriented.

Because of this, GenAI can perpetuate and even amplify discriminatory views in its output. As an instructor, you have the opportunity to help your students recognize and challenge these biases, treating GenAI output as a starting point for critical thinking rather than an authoritative answer.

When you use GenAI, you’re sharing a lot of data in the form of chats and uploaded materials. Under the U of A's agreement with Google, your interactions with Google GenAI tools are not appropriated for model development. Nevertheless, it is advisable to practice good data hygiene in any interaction with GenAI, as it is easy to accidentally include sensitive, proprietary, or private details without considering the risks.

Review the AI Data Safety Guidelines for the University of Alberta for more guidance.

Although many GenAI tools start out free, the best features may be behind a paywall. This can create an unfair learning environment, where some students can afford pro versions while others cannot.

To ensure a level playing field, all members of the U of A community have access to AI tools in the Google Workspace. Be mindful that not all students are equally experienced or open to using GenAI. If you intend to incorporate AI in your teaching, be ready to support student readiness and alternative pathways.

GenAI blurs the lines of who "owns" a piece of work. When you or a student use AI to generate content, the result is a hybrid of human input and machine output. This makes accountability tricky. Some experts say we need to rethink our definition of plagiarism, especially since these models are trained on internet content without the original creators' permission.

In your courses, you can help students navigate this by setting clear rules for how to attribute AI-assisted work. You should also be mindful about uploading copyrighted materials to GenAI, as some publishers forbid doing so.

Visit the U of A Copyright Office offers further guidelines.