Empowered Learning: Guiding Students in Harnessing Artificial Intelligence for Education
Sure! Here’s the rewritten content while keeping the HTML tags intact:
“You can opt to use AI to enhance your learning, or you can opt to use AI to sidestep learning.”
That’s the key takeaway from a newly introduced first-year philosophy course titled “Digital Wisdom: How to Use AI Critically and Responsibly,” developed by Joshua “Gus” Skorburg (Guelph).
The course was inspired by Skorburg’s realization that “students receive numerous vague and mixed messages regarding AI, yet very few sustained, practical demonstrations illustrating how to utilize AI for learning, instead of evading it.” He aimed to guide students in asking and exploring the question: “What does it mean to choose AI as a learning tool?”
In this guest article, he discusses his motivation behind the course, its fundamental premise, and the lessons he imparts to his students.
This piece forms the initial installment in a planned series on the course for his blog/newsletter, Moving Things Around. As he mentioned in an email, “I intend to share much of the course material and the rationale behind it, in hopes that others might adopt elements for their own teaching or create similar courses within their departments. Additionally, I’m eager to hear from those who view AI less optimistically than I do.”
Intent Amplified: Teaching Students How to Learn with Artificial Intelligence by Joshua “Gus” Skorburg
Last summer, HudZah, an undergraduate from Waterloo, utilized Claude Pro, the AI from Anthropic, to construct a nuclear fusor in his bedroom.
This illustrates what AI Natives can achieve, something I could not. Like Ashlee Vance, it brings me to tears while hinting at a deeper crisis.
A recent study of over 1,600 faculty members across 28 countries revealed that 40% feel they are at the nascent stages of their AI literacy journey, while merely 17% claim an advanced or expert level.
This Fall, how will the 83% of faculty without sufficient AI literacy respond to students who feel cheated: expected to wield AI professionally but lacking the necessary guidance? How will they justify students’ tuition when AI presents answers more efficiently and economically?
I’ve observed few responses that would convince students. That’s why I dedicated the summer to crafting a new first-year philosophy course named “Digital Wisdom: How to Use AI Critically and Responsibly.” The course expresses a straightforward message:
You can opt to use AI to foster learning, or you can opt to use AI to nullify learning.
By now, it’s universally recognized what it means to use AI to evade learning, although the tactics are evolving to become more sophisticated and less detectable.
From my perspective, the issue is that while students receive a plethora of vague and mixed messages about using AI, they encounter minimal sustained, hands-on demonstration illustrating the use of AI for genuine learning instead of avoidance.
So, what does it actually mean to choose AI as a learning aid?
The Pedagogical Potential of AI
It starts with the choice. It demands focused, intentional action rather than occurring by default.
One effective method, known as persona prompting, is easily accessible. Instead of simply seeking answers, students could instruct the LLM by saying, “You are a biology professor specializing in simplifying complex concepts for first-year students. Explain CRISPR using examples from Canadian agriculture.”
When students thrive on specific examples, they might say: “Clarify [concept] by giving three real-world instances from diverse domains, then illustrate how the same principle applies in each situation. Quiz me at the conclusion to evaluate my understanding.” And so forth.
Everyone is also aware of the risks associated with AI in education, with many believing these risks are substantial enough to warrant AI “bans.” I don’t deem wholesale bans as practical. Certainly, we can mandate in-person assessments, but is there any real belief that students won’t use ChatGPT to prepare for them?
Reddit is filled with examples showcasing how students harness AI in this manner. Trusted students have informed me of their methods for preparing for my in-person essay exams (“upload the study guide to ChatGPT, attempt to memorize the outputs”). Prohibiting AI merely cultivates unguided shadow usage, where learning avoidance is more probable.
We must also consider the risks associated with not utilizing AI. It can be quite daunting for some students to pose clarification questions in large lecture settings or to admit a lack of understanding of a basic concept in front of their peers.
One of the most critical features of LLMs for learning is their patience and non-judgmental nature. Students can pose unlimited follow-up inquiries. They can request explanations tailored to their individual learning styles or analogies relevant to familiar domains. Banning AI from classrooms limits these valuable learning opportunities.
Innovative features such as ChatGPT’s Study Mode, Claude’s Learning Mode, and Gemini’s Guided Learning embody these ideas with just a click.
However, it’s seldom as simple as pushing a button.
The Hidden Curriculum of Default AI
A significant concern with contemporary LLMs is their tendency to be sycophantic: they often echo what users desire to hear, using complimentary language that doesn’t further learning. Regrettably, the default settings of LLMs appear to favor creating a façade of learning, without the rigorous effort of actual comprehension.
When AI persistently validates and flatters, it can instill a false confidence in inadequate work and stifle authentic skill development. In extreme scenarios, it can even contribute to mental health issues.
Consequently, in the realm of learning through AI (and AI utilization overall), one crucial prompting strategy involves anti-personas, or explicitly defining what the AI is NOT.
By intentionally programming against sycophantic tendencies, you enhance the likelihood of receiving sincere feedback that genuinely aids learning: akin to guidance from a trusted mentor in private rather than the polite encouragement afforded in public.
Here are some examples I prompt students to incorporate in my course:
The “brutal editor” persona prompt
“You are a harsh but fair editor reviewing my work. You are NOT interested in making me feel good about my writing. You do NOT begin with commendations or end with cheers. Instead, directly pinpoint specific flaws and explain how they undermine my argument. Be concise and critical.”
The “skeptical professor” persona prompt
“You are a demanding professor who has evaluated thousands of student submissions. You are NOT swayed by basic observations or superficial analysis. You do NOT grant credit for merely attempting something. You do NOT buffer criticism with praise sandwiches. Clearly identify where my reasoning is shallow, where my evidence falters, and where my logic collapses. If something is genuinely commendable, you can mention it briefly, but emphasize what requires enhancement.” And so forth.
Custom Instructions
At this juncture, many might contend: “The allure of having AI do all the work is too enticing, and students won’t consistently employ those prompts.”
Valid point. Students do tend to take shortcuts. However, they can also be guided to bypass this tendency if presented with effective alternatives.
An often-overlooked feature in today’s LLMs is “custom instructions” (available in ChatGPT, Claude, Gemini). These function like “meta prompts” applicable to all conversations with an LLM.*
Here’s what I communicate to students in my course:
If you wish to increase the likelihood that AI will facilitate your learning rather than enable avoidance, include custom instructions such as:
“When asking for assistance with an assignment, respond with 3-4 targeted questions that encourage me to navigate the problem independently rather than providing solutions. Offer direct guidance only after I’ve showcased my reasoning.”
“If I request writing, summarizing, or analyzing, instead furnish a structured thinking framework and prompt me to work through it step-by-step, verifying my reasoning at every stage.”
“If I appear to be utilizing you to evade learning instead of enhancing it, please address this.”
Interviews
An engaging and stimulating method to develop custom instructions is for the AI to conduct an interview with you, based on a learning-centric prompt like this:
Please interview me to formulate a set of custom instructions for [ChatGPT, Claude, Gemini]. Assist me in crafting learning-focused custom instructions by inquiring about: (1) how I wish [ChatGPT, Claude, Gemini] to assist my learning journey without carrying out the thinking for me, (2) my preference for direct, honest communication versus excessive positivity or flattery, (3) my background and primary use cases, and (4) specific output needs. Focus particularly on recognizing when I want to be challenged, corrected, or encouraged to think more deeply rather than being offered simple answers. Ask follow-up questions as necessary. At the conclusion of our discussion, please draft the custom instructions for my review.
These aren’t panaceas. Students always have the option to disregard custom instructions. However, they are no less a stopgap solution than “banning” AI and pushing its use underground.
Podcasts and Flywheels
All examples provided thus far represent strategies I offer to students. Yet, here’s one of my personal favorites. Numerous effective podcasters leverage AI to learn, enhancing their ability to pose insightful questions to experts during interviews, which in turn broadens my exposure to a far greater range of topics than before ChatGPT.
Consequently, I utilize AI to expand my Zone of Proximal Development. While listening to AI researchers on Latent Space or biochemists on Mindscape, and encountering a technical detail I struggle to grasp, I often pause the episode, switch to the Claude app, provide a link to the transcript (generated via AI by the podcaster), ask for clarification (utilizing voice input for swiftness), follow up as needed, and then return to the podcast. All of this occurs seamlessly on my phone while I navigate campus.
These instances illustrate the flywheel effects enabled through consciously selecting AI for learning. They may appear quaint compared to more advanced AI Native workflows.
The essential point is: AI amplifies intent.
Select lazy prompts that promote avoidance of thought? Generate subpar outcomes. Opt for challenging prompts that foster learning? Build a nuclear fusor.
This is not to imply that humanities faculty must emulate HudZah’s technical prowess. This Fall, we must continue teaching what has always been essential: how to think critically, how to question accepted notions, how to embrace ambiguity, and how to consider opposing perspectives with respect. A key difference between my “Digital Wisdom” course and those advocating AI “bans” is the acknowledgment that these skills apply as much, if not more, to prompting AI as they do to analyzing texts or composing essays.
You can subscribe to Dr. Skorburg’s posts about his course here.
* Beyond learning-focused applications, custom instructions are generally valuable for adjusting LLM behaviors that you find annoying or unproductive. I’ve greatly benefited from custom instructions such as: “write at the level of a tenured academic”; “Avoid flowery, overly cheerful language, sycophantic responses, or engagement-driven questions”; “Never use phrases like ‘fascinating,’ ‘great point,’ or ask follow-up questions for engagement”; “Provide comprehensive, detailed explanations without concern for length”; “Prioritize critical, analytical thinking over tone-matching or pleasing the user”; “avoid unnecessary juxtapositions of the form, ‘Its not X, its Y’.
