Navigate This Resource

What is Generative AI?

Institutional Policies

AI in Academic Practice

Building Your Skillset

Ethical Implications

Pedagogical Considerations

Cultivating AI Literacy

AI Futures

On vetting AI tools: Choosing where to spend your time and data

As AI tools become increasingly integrated into academic and professional work, it’s important to go beyond surface-level functionality and assess them critically. Not all tools are equally transparent, ethical, or effective — and how you use them matters just as much as what they produce.

The University of Toronto developed VALID-AI framework offers a structured way to evaluate AI tools and your own engagement with them. Use it to guide your decisions about where to invest your time, how to interpret results, and what ethical considerations to keep in mind when working with AI in your research, writing, and learning. Below you will find the framework with additional questions and ideas for executing each step of the vetting process.

V: Validate data – Assess the quality and reliability of the data used to train and test the AI model. Is the data representative, unbiased, and relevant to the problem at hand?

How can I do this?

Look for “model cards” or help documentation — does the tool explain what data it was trained on?
Test it using diverse topics or identities. Does it respond equitably?
Notice what’s missing — are some perspectives excluded or misrepresented?
Cross-check output with academic sources, not just surface-level web results.

A: Analyze algorithms – Examine the AI model’s underlying logic and learning processes. How much human-reinforced learning is present? Could it introduce bias?

How can I do this?

Look for whether the model was trained with human feedback — this can shape how “neutral” it is.
Try prompts from different epistemological or cultural viewpoints and compare responses.
Be cautious of overconfident outputs — bias may be masked by fluency.

L: Legal and ethical considerations – Evaluate whether the AI tool complies with relevant legal and ethical standards. Does the tool respect privacy, consent, and responsible data use?

How can I do this?

Check whether the tool stores or reuses your data — especially uploaded documents.
Read the privacy policy or FAQ, even briefly. If it’s unclear, proceed cautiously.
Don’t use tools that require you to log in with sensitive academic accounts unless officially approved.

I: Interpret how it works – Assess whether the AI’s outputs are explainable and transparent. Is it clear how the AI generates its responses?

How can I do this?

Ask the tool to explain its reasoning (“How did you get that answer?”).
Note when it refuses a prompt — does it explain why?
Use tools that cite sources or outline their steps, when available.

D: Diversity and bias – Evaluate whether the AI reflects a broad and inclusive range of perspectives. Does it reinforce stereotypes or favour dominant frameworks?

How can I do this?

Try prompts using varied cultural, linguistic, or social lenses — see how the tool responds.
Notice what it assumes as “neutral” or “normal.” Who gets centred?
Reflect on who benefits and who might be harmed by the outputs.

A: Accuracy check – Cross-check AI output against trusted sources. Is it verifiable and accurate, or vague and misleading?

How can I do this?

Ask: “Where is this information from?” If there’s no citation, treat with caution.
Always verify citations.
Compare results to academic journals, textbooks, or instructors’ explanations.
If it gives made-up citations (hallucinations), proceed with caution.

I: Intent and impact – Reflect on your purpose and the potential effects of your AI use. How does your use of AI affect others — including human creators?

How can I do this?

Ask yourself: “Why am I using this tool? Is it helping me think, or replacing the thinking?”
Be transparent if you’re submitting work that included AI support.
Compose an AI Use Statement that is thoughtful and reflective.
Keep revisiting this question.

Developing your AI Skillset

AI Skill Domains

Prompting	Develop precise, useful, explanatory prompts that trigger meaningful outputs.
Reading	Prompt design, Critical Evaluation, Fact Checking, Perspective Analysis, Dialogic Inquiry, Reflective Awareness
Learning	Goal-Setting, Asking Clarifying Questions, Identifying Misconceptions, Iterative Practice, Self-Assessment, Reflective Thinking, Verifying Accuracy, Customizing Feedback
Writing	Idea Generation, Prompt Refinement, Draft Development, Voice and Tone Adjustment, Structure and Organization, Revision Strategy, Citation and Attribution, Plagiarism Awareness
Image Creation	Prompt Crafting, Iterative Refinement, Ethical Image Use, Cultural Sensitivity
Vibe Coding	Tinker and build with AI-generated code, DIY coding

🛠️ Prompt Engineering: General Purpose Tools

Prompt engineering is the practice of crafting thoughtful and intentional inputs to guide an AI system toward producing useful, relevant, or specific outputs. Prompts act as the bridge between you and the LLM — the way you ask a question or structure a task directly shapes how the model responds. Effective prompts involves paying attention to wording, context, and your purpose, so that the AI can better understand what you're asking and respond in a meaningful way.

It is important to remember that LLMs are designed to be agreeable and can often sacrifice truthfulness or validity in favour of telling you what you want to hear (for more on this topic read more about the problem of hyper Sycophancy and listen to this excellent podcast on the way these models are designed: Echo Chambers of One .

For this reason it can be helpful to use a prompt that tells the LLM how you want it to respond.

Many people have developed elaborate and carefully crafted prompts that can be used as a framework for your searches. One excellent framework is the Sensemaking with SIFT toolbox, developed by Mike Caulfield.

⚡ Check the validity of a claim by referring to research

This prompt can be used to fact-check the validity of a claim such as: “The environmental impact of using LLMs in one’s workflow is a growing cause for alarm.”

First, the SIFT toolbox prompt was entered into Claude.ai, then the claim was entered. I then directed Claude to follow the instructions and review the claim accordingly.

🔍 Click to preview only a part of the output:

Generated May 15, 2025 – may be out of date if significantly later.
AI-Generated: Will likely contain errors; treat this as one input into a human-checked process.

📗 Revised Summary (Corrected & Accurate)

Large language models do have a measurable environmental impact, primarily through energy consumption and water usage for cooling data centres. Training a model like GPT-3 produces approximately 552 metric tons of CO₂ — equivalent to the emissions from 123 cars driven for a year (Scientific American).

Individual LLM usage has a relatively small environmental footprint — about 2–4g CO₂ per query, compared to 0.2g for a Google search (Smartly.AI). Even frequent use (10 queries daily for a year) equals just 0.07% of an average American’s carbon footprint (Sustainability by Numbers).

However, the aggregate impact becomes substantial: with 164 million monthly users, ChatGPT could collectively generate emissions equivalent to 260 transatlantic flights monthly (Sustainability News). Paradoxically, AI may be more efficient than humans for certain tasks — recent research shows AI can produce 130–1500x less CO₂ than human-generated content (Nature).

🏅 What a Fact-Checker Might Say

Partly True, Needs Context. While LLMs do consume energy and water, individual use has minimal environmental impact — much less than daily activities like driving or heating. The greater concern lies in aggregate industrial use as LLMs scale globally. In some domains, AI may be more carbon-efficient than humans.

💡 Tip Suggestion

When evaluating claims about tech and the environment, separate individual vs. collective impact and compare emissions to familiar benchmarks. For example, boiling a kettle releases ~70g CO₂, much more than a single LLM query.

There are many tools out there for building and sharing great prompts. Take a look at OpenAI's own GPT-4.1 Prompting Guide for how to build useful prompts using Chat GPT.

In general, you can think about any prompt you give an LLM on a spectrum between open-ended and highly specific. Remember that the more open-ended the prompt is, the more you leave up to the tool (and the designers of the tool). It is important to remember their implicit, baked-in biases, including: frequency bias, agreeability bias, gender and racial bias.

Some extra resource for guidelines on crafting prompts:

Harvard University (2025). Getting started with prompts for text-based Generative AI tools. https://www.huit.harvard.edu/news/ai-prompts
Akin, F. K. (Ed.) (n.d.) prompts.chat. World's first & most famous prompts directory. https://prompts.chat/
University of Michigan (n. d.). Prompt literacy in academics. https://genai.umich.edu/resources/prompt-literacy

⬆️ Back to List of Domains

📘 Reading with AI

The best way to read a great article, book, chapter, or any text you want to engage with is to do just that—read it. Beginning to end, top to bottom, quietly and thoughtfully—maybe even more than once. This kind of immersive reading is foundational to graduate study and should be protected in both your academic work and your life.

However, deep reading isn’t always feasible—or necessary—for every task. Throughout your graduate studies, you’ll encounter many types of reading demands, from surface-level scans to close, analytical engagement. AI tools can support your reading practices in several helpful ways, particularly when used thoughtfully and strategically:

Identifying Key Texts: When navigating large volumes of research—journal articles, books, websites—AI can help you synthesize content and identify which texts or sections warrant closer, full-length reading.
Decoding Difficult Texts: If you encounter complex or dense material, AI can offer clarifications, summaries, or simplified explanations to support your understanding without diluting the core ideas.
Supporting Active Engagement: AI tools can also help you engage more critically by generating reflective questions, offering organizational aids, or helping you test your comprehension of a text.

Reading involves critical analysis, conceptual synthesis, and making original connections across ideas and texts. Used judiciously, AI can enhance—but should never (never!) replace—deep reading and reflective thought.

Prompts to Try

For Critical Engagement
How can I harness AI to help me develop deeper interpretive and analytical practices as I read academic texts?

How does the author define the core concept, and how does it differ from other definitions?
Explore how the author defines key terms. Compare this to other foundational or emerging texts in the field. What’s at stake in these differences?
What assumptions underlie this argument?
Identify what is taken for granted—about learners, technology, systems, or knowledge. Consider whose voices are missing or marginalized.
What disciplinary discourse is the author situated in?
Consider which academic conversations the text participates in or pushes back against. Is it aligned with instructional design, critical pedagogy, learning sciences, etc.?
Are there implicit values or epistemologies shaping the position?
Notice whether the argument reflects a particular worldview or way of knowing—positivist, constructivist, decolonial, etc.—and what that means for interpretation.
What contradictions or ambiguities are present?
Pay attention to tension points, paradoxes, or ambiguity in the author’s claims. What opportunities for critical reflection or further inquiry do these open up?

For Conceptual Mapping & Discourse Navigation
How is this text positioned within broader conversations in the field, and what relationships, tensions, or patterns connect it to other thinkers, texts, or traditions?

What other authors or texts does this article reference or respond to, and how does it position itself in relation to them?
Can you identify the key concepts in this article and describe how they build on, contrast with, or complicate those in [insert reading or author]?
How does this author build on, revise, or challenge dominant discourses in [insert field or tradition, e.g., educational technology or critical pedagogy]?
Show how the concept of “[insert term]” is framed in this article versus how it’s used in another key text we’ve read.
What academic conversations (e.g., critical theory, design-based learning, AI in education) is this article participating in or pushing back against?
I’ve uploaded several articles—can you compare how each one frames [insert concept] and highlight key similarities, differences, or shifts in perspective?

For Synthesis & Strategic Reading
How can AI help me practice and scaffold high-level academic reading skills?

How does this author’s argument compare to [Author X]'s ideas about [concept]? What do they agree or disagree on?
Summarize the methodological approach in this article. How does it differ from [another method or study]?
What is the intellectual history of the concept “[insert concept]”? Where did it originate, and how has it evolved across fields?
Reinterpret this article through a [feminist / Indigenous / posthumanist] lens. What new insights or critiques emerge?
This is what I’m interested in: [insert your thesis, question, or topic]. Can you direct me to the most relevant sections of this text to focus my reading?
Given my interest in [insert concept], which parts of this article should I prioritize, and why?
Can you outline the structure of this article and highlight the sections most relevant to [insert research goal or assignment prompt]?
Summarize this article in 3–5 sentences, focusing on the main argument and how it contributes to the field of educational technology.
What does this article contribute that might be useful for someone researching [insert topic]?

Tools to explore

Notebook LM
Google Scholar's Secret AI Feature
Elicit: The AI Research Assistant (check out this Product Review)
Scholarcy

⚠️ Caution: Use these tools mindfully to support—not replace—your learning: See Letting AI read for us can undermine our thinking. (Baron, 2024).

⬆️ Back to List of Domains

🧠 Learning with AI

In the MET program, you’re constantly learning across a range of domains. Whether you’re trying to understand sociocultural learning theory, prototype a digital activity, evaluate an edtech tool, or craft a strong academic analysis, your ability to pause and reflect on how you’re learning is just as important as what you’re learning. This kind of metacognitive awareness—asking yourself what you’re trying to grasp, where you’re getting stuck, and what kind of support you actually need—is a core skill in graduate-level inquiry. When used with intention, AI can support and extend this reflective process by helping you:

Rehearse or rephrase an idea to see if you truly understand it
Explain a concept in a different way (e.g., simpler terms, metaphors, or examples)
Test a design or approach against pedagogical or learning goals
Identify gaps, contradictions, or weak logic in your developing argument or plan

But AI can’t (and shouldn’t) do the learning for you. It becomes most valuable when it responds to your thinking—not when it replaces it. Let it be a thoughtful companion in your learning journey—not a shortcut around it.

Prompts to try

Check Your Understanding

Here’s how I understand [insert your explanation of the concept]. Can you restate it in a different way so I can check my understanding?
Can you help me put this idea [insert a full description of your idea] into simpler or clearer language, without losing its meaning?
Please interview me on a given topic. Ask me questions that check the depth and breadth of my understanding.
Can you quiz me on [insert information]?

Explain a Concept in a Different Way

Explain [insert concept] using a metaphor or real-world example relevant to [insert area of interest].
Can you explain [insert theory] as if I were a new MET student unfamiliar with this framework?

Test a Design or Approach Against Learning Goals

I’m designing an activity to support [insert learning goal]. Based on what I’ve described, does this design align with that outcome?
What are some potential gaps or misalignments between this design and constructivist learning principles?

Identify Gaps, Contradictions, or Weak Logic

Here’s my working argument for an essay: [insert argument]. Can you identify any gaps, contradictions, or areas where it could be stronger?
What are some possible counterarguments or limitations to this idea that I should consider?

⚠️ Caution: Remember to always check the Syllabus AI Use Statement to ensure you are allowed to use AI tools to help you with a course-related task. If you are in doubt, make sure to contact your instructor.

⬆️ Back to List of Domains

✍️ Writing with AI

For many educators, writing is the area of greatest concern when it comes to student use of AI — and for good reason. Writing is not just about output; it's a personal, often difficult process through which thinking develops. As Joan Didion famously put it, “I don’t know what I think until I write it down.” If AI does the writing for you, that discovery is lost. But used thoughtfully, AI can support the writing process — not replace it. With effort and reflection, it can act as a scaffold, offering feedback, revision prompts, and opportunities to clarify and sharpen ideas. Some ways to use AI in your writing process are:

Generate multiple drafts of an idea to experiment with voice, structure, and tone
Identify unclear or awkward phrasing and improve readability
Simulate feedback to help you revise, expand, or reframe your writing
Help you structure your thoughts into a logical flow—such as outlining an argument, identifying key sections, or sequencing points for clarity and progression.
Practice synthesizing and paraphrasing, so you can integrate source material more effectively
Explore counterarguments or ethical implications you may not have considered

⚠️ Caution: These tools are most powerful when they are in dialogue with your thinking — not acting in place of it.

For Developing a Thesis

"Assume the role of a thesis advisor with a specialization in academic writing. Your task is to refine and enhance a student's thesis statement for their academic paper. The thesis statement provided by the student is: "[INITIAL THESIS STATEMENT]." Your goal is to critique this initial thesis statement for clarity, specificity, and academic rigor. Provide constructive feedback on how to make the thesis more concise, focused, and argumentative. Suggest ways to incorporate a clear stance, significant research questions, and the potential implications of the research. Ensure the refined thesis statement is well-defined, debatable, and sets a solid foundation for an interesting argument in their academic paper."

from: https://www.godofprompt.ai/prompt-library
"As a research advisor, your task is to brainstorm potential research topics for an academic writing project in "[FIELD OF STUDY]." Begin by identifying current trends, gaps, and emerging questions within this field that could lead to impactful research. Consider interdisciplinary approaches that could enrich the research perspective. Generate a list of at least five interesting research topics, each accompanied by a brief explanation of its significance, potential research questions, and the contribution to the field. Ensure that each topic is workable, original, and has a clear scope for in-depth academic investigation."

from: https://www.godofprompt.ai/prompt-library

Explore and Expand Your Thinking

I’m writing about [insert topic]. Can you help me brainstorm different angles or critical questions I could pursue?
What are some emerging debates or unresolved issues related to [insert topic] that I could explore?
Can you give me a range of theoretical lenses I might use to approach this topic?
What kinds of examples or case studies would help ground this concept?

Draft and Experiment with Voice

Here’s a rough paragraph I’ve written: [insert text]. Can you rewrite it in a more formal academic tone?
Can you adjust this writing to sound more critical, reflective, or exploratory?
How would this argument change if it were framed as a narrative or a provocation?
What’s a more concise way of making this point, while keeping the nuance?

Organize Ideas into a Scaffolded Framework

Here’s my research topic: [insert topic]. Can you help me outline a scaffolded structure for a graduate-level paper, including section headings and subpoints?
These are the points I want to make: [list points]. Can you arrange them into a logical order and suggest how I might transition between them?
Can you map my argument using an intro-body-conclusion structure, and show where I might need evidence, definitions, or counterpoints?
Based on my thesis—[insert thesis]—what 3–4 main sections could I use to support it?

Clarify and Refine Language

Can you help me rewrite this sentence for clarity without changing its meaning?
Are there any overly complex or awkward phrases in this paragraph?
What parts of this passage might confuse a reader? Can you offer suggestions to improve flow?
Highlight where I need to define terms or provide clearer transitions between ideas.

Reflect and Revise Critically

I’m writing about [insert topic]. Can you help me brainstorm different angles or critical questions I could pursue?
What are some emerging debates or unresolved issues related to [insert topic] that I could explore?
Can you give me a range of theoretical lenses I might use to approach this topic?
What kinds of examples or case studies would help ground this concept?

Draft and Experiment with Voice

Here’s my working argument: [insert argument]. What are some possible counterarguments or limitations I should address?
Can you identify any logical gaps, contradictions, or weak transitions in this section?
I’ve revised this paragraph—can you compare it to the original and tell me which version communicates better and why?
What implicit assumptions am I making in this argument that I may need to name or challenge?

⬆️ Back to List of Domains

🎨 Creating images with AI

As with writing, the features of AI-generated images are shaped by the ingredients of your prompts, the tool itself, (along with all the human creators of the artworks and images the model has been trained on!). In this section we will explore the different ingredients you can add to your prompt in order to increase your control over the output.

⚠️ Cautionary Note: Copyright and AI-Generated Images

Click to learn more about legal and ethical landscape around copyright and authorship.

Remember to consider the following as you explore using and creating with tools like DALL·E, Midjourney, or Stability AI:

Training Data May Include Copyrighted Material: Many AI image generators have been trained on vast datasets scraped from the internet—including copyrighted artworks—without the original creators’ consent. This is at the heart of several ongoing lawsuits, including Andersen v. Stability AI and cases against Google’s Imagen.
You May Not Own the Copyright: In many jurisdictions, works generated solely by AI without significant human input are not eligible for copyright protection. This limits how you can claim or license such images.
Style Mimicry Is a Grey Area: Generating images “in the style of” specific artists raises ethical and legal concerns. Even if not illegal, this can violate professional norms or terms of use.
Be Transparent and Respectful: If using AI-generated imagery:
- Credit the tool used
- Acknowledge limitations or inspirations
- Avoid presenting work as wholly original unless significantly altered or curated

Why It Matters: AI-generated images offer exciting creative opportunities—but they also challenge our assumptions about authorship, labor, and ownership. Responsible use means staying informed and critically engaged with the ethical terrain of digital creation.

Like all prompt design, you can start wide open — for example, cyborg— or hone your instructions with more precision and details you would like to see in the image. The more detail you provide, the more specific and aligned the results will be with your desired output. Remember to always take a critical eye to the output: What default imagery does the model use? Whose visual cultures are being privileged or erased? What visual clichés or tropes does the model rely on?

As you view the following images of the "cyborg," consider that they are not neutral or universal representations. Instead, they draw on visual tropes and cultural assumptions that have long shaped speculative imagery in the global North—often reflecting gendered, racialized, and ableist ideas about bodies and technology. What perspectives are being amplified here? Which ones are missing?

Midjourney output with the entry: Cyborg

Prompts can include elements like: subject, context, style, mood, composition, material, lighting, or perspective. As with all prompts, being more specific can direct the output to align more closely with your idea.

Midjourney and ChatGPT Output with the Following Prompts:

ChatGPT Prompt:
Can you give me a prompt including these elements: subject, context, style, mood, composition, material, lighting, or perspective -- around the theme of a cyborg AI

Midjourney Prompt:
A solitary cyborg AI standing at the edge of a rusted, post-human city at dusk, partially covered in vines and moss, suggesting time and decay. The cyborg's design is sleek but visibly repaired with mismatched components — a fusion of organic tissue and retrofitted metal plates. Style is reminiscent of Moebius and late-70s sci-fi concept art. The mood is contemplative and melancholic, evoking a sense of fading memory and machine loneliness. The composition is wide-angle, with the cyborg in the foreground gazing over a broken skyline, bathed in diffused amber light. Materials include chrome, ceramic, bark-like textures, and faded synthetic fabric. The lighting is soft and cinematic, with long shadows and a hazy atmosphere. Perspective is from behind the cyborg, slightly above shoulder height, drawing the viewer into the scene.

Just as with generative AI for text, you can steer and guide image-generation models by uploading reference or guide images that influence the visual style, composition, or content of the output.

While many visual artists are concern with the impacts of AI on artistic practice, some contemporary artists have embraced AI into their process. Check out Argentine artist Sofia Crespo who uses AI models trained on datasets of natural organisms, often incorporating her own photographs as reference images. This method enables her to create intricate, nature-inspired artworks that question the boundaries between the organic and the artificial.

Image source: https://nextnature.org/en/magazine/story/2020/interview-sofia-crespo

There are many resources available for you to explore. Check out the following as a starting place:

⬆️ Back to List of Domains

💻 Vibe coding with AI

Screenshot 2025-05-20 at 11.51.11 AM.png

Vibe coding is an emerging approach to software development in which programmers and non-programmers alike use natural language prompts to guide large language models (LLMs) in generating code, shifting the design/building process from manual coding to steering, testing, and refining AI output.

Popularized by Andrej Karpathy in early 2025, the term captures a more intuitive, conversational mode of programming—"seeing things, saying things, running things"—that lowers barriers for amateur developers but often sacrifices deep understanding of the code.

Unlike traditional software engineering, vibe coding embraces a form of interaction where users may accept AI-generated solutions without fully grasping their inner workings. While Karpathy sees it as useful for rapid prototyping and weekend projects, critics argue that true vibe coding involves relinquishing comprehension, distinguishing it from more deliberate uses of AI as a coding assistant.

While there are valid critiques of relying on vibe coding, it can significantly lower the barrier to entry for creating interactive apps, media, and websites. By making these tools more accessible to non-programmers, vibe coding can help jumpstart the learning process and support rapid prototyping.

A growing number of dedicated platforms now support this approach—for example, v0 by Vercel, Bolt AI, and Replit all provide intuitive environments for building projects with minimal coding experience. If you have an idea for an app, these tools make it easy to dive in and start developing a prototype.

You can also use general-purpose large language models (LLMs) to generate code snippets for your projects, even if you're just beginning to learn.

As vibe coding becomes more common, so do the resources for getting started. Just be sure to consult your instructor if you're planning to use this approach for a course assignment or project.

✏️ ✏️ Course developer aside:

Below is the ChatGPT exchange I used to code this callout box. I simply copied and pasted the code into the html
editor to create the layout. I built other interactive media that you will find in this resource using ChatGPT and https://codepen.io/ to test it out live online.

Screenshot 2025-05-19 at 9.27.41 AM.png

⬆️ Back to List of Domains