✨Guide: How to Train Students to Use AI

Welcome to AutomatED: the newsletter on how to teach better with tech.

Each week, I share what I have learned — and am learning — about AI and tech in the university classroom. What works, what doesn't, and why.

In this fortnight’s Premium edition, I present a guide to help you train your students to use AI, from adjusting and adding AI-relevant course objectives to helping students prompt better and develop AI literacy.

🔎 Step 1: Assess the Role of AI in Your Field

Learn AI and Take Stock

The first thing to consider is very “big picture” and abstract, but it is essential.

You need to start by reflecting on the extent to which AI can augment or enhance the core tasks, processes, and outputs of your field. This assessment will help you determine the depth and breadth of AI training needed in your course.

To evaluate your field's AI potential, consider the following questions:

1. What are the primary tasks, processes, and outputs in your field that could be completed entirely by AI?

2. How might AI tools enhance or streamline tasks, processes, and outputs in your discipline?

3. What new possibilities or areas of exploration could AI open up in your field?

When reflecting on these questions, it is important to think broadly about the nature of your discipline and its methodologies. If applicable, consider both the quantitative and qualitative aspects of your field. For quantitative tasks, AI might offer advantages in data pre-processing, pattern recognition, predictive modeling, and visual representation. For qualitative tasks, AI could assist with text analysis, critique or justification, and ideation.

Crucially, you need to be forward-looking; try to imagine how AI might transform your discipline in the future. If you can see how an AI tool, if improved 10%, would be relevant or a game-changer, then reflect on how you could prepare your students to use it once it improves. Probably, you won’t have to hold your breath long for the improvements to be realized.

Relatedly, for many fields, it is essential to consider the future demand for a certain skill in the job market. If the industry is moving towards AI-assisted tools and technologies, teaching the use of current versions of the tools can provide students with more relevant skills for their future careers, even if the tools aren’t quite ready for full-time use.

Now, to effectively conduct this reflection on AI’s relevance to your field, it's crucial to first develop a solid understanding of AI's capabilities. This can be achieved through two main approaches: hands-on experimentation with AI tools and gathering insights from external sources.

Experimenting with AI tools yourself is particularly valuable. Given the unique nature of your teaching style and course content, you're best positioned to evaluate how AI can enhance your specific pedagogical approach. What works for one professor may not suit you, making personal exploration essential.

When experimenting, commit to extended engagement with each tool. AI tools often have a learning curve, and only through sustained use can you fully grasp their potential. Some claim that it takes at least 10 hours with a tool to learn how to use it to a minimal standard.

Remember: while your expertise enables you to judge the quality of AI outputs, developing proficiency in eliciting optimal results from these tools takes time and practice.

Here are some examples of tools that I would recommend experimenting with, organized by category…

Large language models:

• GPT-4o - OpenAI's most advanced model, known for its high accuracy and versatility across various applications such as text generation and reasoning tasks​.

• Claude 3.5 Sonnet - Developed by Anthropic, this model may have recently surpassed GPT-4o in some dimensions, but the consensus is that it is comparable​ to GPT-4o while being a better writer.

• Gemini 1.5 - Google DeepMind's latest model, notable for its massive context window and comparable performance in multiple benchmarks​.

• Poe - A way to interface with and compare a range of LLMs.

Large language models paired:

• ChatGPT’s Advanced Data Analysis - This tool combines the capabilities of GPT-4 with advanced data analysis features.

• ChatGPT’s Custom GPTs - Allows for creating tailored models suited to specific needs, enhancing the versatility of large language models.

Code assistants:

• The LLMs listed above

• GitHub Copilot - Developed through a collaboration between GitHub, OpenAI, and Microsoft, Copilot is renowned for its code suggestions, autocompletion, and documentation insights. It supports a wide range of programming languages and integrates seamlessly with popular IDEs such as Visual Studio Code, JetBrains IDEs, and more.

• Amazon CodeWhisperer - This tool supports multiple programming languages including Python, Java, JavaScript, and C#. It offers real-time code suggestions and is optimized for use with AWS services.

• Tabnine - A versatile AI coding assistant that provides code creation, explanations, testing, documentation, and bug fixing. It supports various programming languages and integrates with popular code editors, enhancing productivity and code quality​.

Visual design tools:

• Canva - Stands out for its user-friendly interface and extensive library of templates, making professional-quality design accessible to users of all skill levels, with powerful AI tools for tasks like background removal and text generation​.

• Adobe Firefly - Integrates advanced AI features directly into Adobe Creative Cloud applications, allowing for seamless text-to-image generation and enhanced productivity within professional design workflows​.

Image generators:

• DALL-E 3 - Ease of use and high fidelity in translating detailed text prompts into precise, contextually rich images. Accessible through integration with ChatGPT​.

• Midjourney - Able to generate highly artistic and imaginative images, making it a favorite among artists and designers who seek creative and visually striking outputs​

• Stable Diffusion - Offers extensive customization and control through open-source availability, enabling users to run and train the model on their own data, which is ideal for those seeking deep customization and photorealistic outputs

Research productivity tools:

• Perplexity - Perplexity stands out for its advanced AI-driven research assistance, delivering quick, detailed answers to complex questions and supporting researchers with high-quality, synthesized information from various sources

• Superhuman - Improved email productivity by leveraging AI to automatically sort and prioritize emails, and extensive keyboard shortcuts to speed up email management.

• ResearchRabbit - Specifically tailored for academic researchers, providing AI-driven recommendation, summarization, and visualization features, making it ideal for efficiently finding, organizing, and citing academic papers​ and authors.

Yet, while hands-on experimentation is invaluable, it has its limitations. AI is one of the fastest growing areas of scholarship and practice, after all.

Thus, it is also essential to efficiently gather insights from experts and experienced users about various AI tools' capabilities. Complementing your hands-on experience with external perspectives — through tutorials, analyses, videos, courses, and discussions — can provide a well-rounded understanding of AI's educational applications.

I recommend adopting a 'sampling' approach to acquire these insights. This method involves strategically exploring the diverse landscape of AI information sources, focusing on those that address issues relevant to your teaching practices. Once you've identified valuable sources, maintain regular check-ins to stay abreast of significant updates or developments. When necessary, conduct deeper investigations into their past content. As you encounter references to other potentially useful sources, explore those as well, gradually expanding your information network.

This process is analogous to the research phase of academic writing. Most of us skim numerous books, journal articles, and encyclopedia entries, searching for threads that related to our paper's thesis. We then go deeper when relevance is sufficiently likely.

Similarly, in exploring AI resources, you're seeking information that resonates with your pedagogical needs and interests. This targeted yet flexible approach allows you to efficiently navigate the expansive world of AI, curating a personalized knowledge base about AI’s potential role in your field that informs your teaching strategies.

Fundamentals as Prerequisites?

With that said, it's also crucial to consider the skills and knowledge that are fundamental to your field. How might AI interact with these core competencies?

In some cases, AI might augment these skills, allowing students to apply them at a higher level or in more complex scenarios — but only if they already know what to look for (i.e. what good outputs or performances look like). In other cases, AI might automate certain tasks with less supervision needed, potentially freeing up time for deeper analysis or more advanced topics.

While AI tools can undoubtedly enhance many aspects of learning and professional practice, they should not supplant the development of fundamental skills and knowledge that are essential to your discipline, including those essential to skillfully using AI itself.

Consider the following points:

1. What core competencies in your field are essential for students to develop, regardless of AI's capabilities?

2. How might these fundamental skills and knowledge enable students to better utilize and evaluate AI tools?

3. In what ways could an overreliance on AI potentially hinder students' deeper understanding of your subject matter?

It's tempting to embrace the notion that "AI is the future" and focus solely on training students to use AI tools effectively without enough time spent on developing core competencies. However, this approach risks creating a dependency that could ultimately limit students' abilities.

Just as a student who consistently copies a classmate's work fails to develop the judgment needed to evaluate that work's quality, a student overly reliant on AI may struggle to critically assess AI-generated outputs.

The goal, therefore, is to strike a balance. While incorporating AI training into your curriculum, you may also want to design "AI-immune" assignments that focus on developing the fundamental skills and knowledge of your field. These assignments serve a dual purpose:

1. They ensure students acquire the core competencies of your discipline, maintaining the integrity and depth of their education.

2. They equip students with the judgment and critical thinking skills necessary to effectively use and evaluate AI tools in your field.

By isolating what students need to know and be able to do without AI assistance, you're actually preparing them to be more discerning and capable users of AI in the long run.

As you progress through this Guide, keep in mind this balance between embracing AI's potential and preserving the development of core disciplinary competencies. This balance will inform your decisions about how and when to integrate AI training into your curriculum.

🛠️ Step 2: Adjust Existing Course Objectives

Review Existing Course Objectives

After you have a strong grip on the role of AI in your field and teaching more generally — as well as to what extent you need to develop your students’ fundamentals as prerequisites to effective AI use — you need to connect these insights to the specific course you are teaching.

One reliable way to do this is via a reflection on your course’s learning objectives. (This method is called “backwards course design.”)

Begin by thoroughly reviewing your existing course learning objectives, assuming you’ve taught this class or one like it before. If you haven’t taught this class or one like it before, then try to express the course objectives you would have developed before the advent of AI. These objectives typically encompass knowledge acquisition, skill development, and the cultivation of specific competencies within your field.

Examine each objective critically, considering its potential relationship to the AI tools and methodologies that you decided above to be relevant to your field and your students.

Adjust Objectives Based on AI Relevance

As you review your existing objectives, consider categorizing them based on the degree to which AI will affect them.

And remember: as you categorize each objective, consider not just its current relationship to AI, but also how it may need to reflect how AI technologies will evolve in the coming years. This forward-thinking approach will help ensure that your course remains relevant as your field develops over time.

In short: many of your students won’t be coming back to school again after this degree!

One way to do this is to create three categories:

1. Existing objectives that could be significantly enhanced by AI integration

For objectives that could be significantly enhanced by AI integration, think about how AI tools might fundamentally change the way students engage with the material or develop skills. These might be objectives that involve complex data analysis, pattern recognition, or tasks that can be greatly accelerated or expanded in scope through AI assistance.

Consider questions like the following:

• How might AI tools facilitate deeper understanding of key concepts?

• In what ways could AI assist in skill development or application?

• Could AI provide new perspectives or approaches to problem-solving within your field?

• Might AI offer opportunities for more advanced or complex explorations of course material?

2. Existing objectives that might be moderately impacted or supported by AI

In contrast, objectives that might be moderately impacted by AI could include those where AI serves as a supportive tool, enhancing the learning process without fundamentally altering it.

3. Existing objectives that should remain independent of AI influence

Finally, identifying objectives that should remain independent of AI influence is crucial for preserving those prerequisites that students must understand or meet, regardless of whether they will be deploying AI later on. These might include objectives related to evaluating the quality of outputs in your field (e.g. excellent presentations, argumentative essays, or analytic reports) or objectives focused on ethical reasoning, creative expression, or the development of interpersonal skills.

Consider scenarios where reliance on AI might shortcut important cognitive processes or practical skills that students need to develop. While an AI tool might be able to quickly generate a solution to a complex problem, the process of grappling with that problem manually might be crucial for developing a deep understanding of the underlying principles. (This goes back to the question of fundamentals from section 1.)

Identify these potential conflicts by asking yourself questions like the following:

• Could the use of AI in this context bypass important cognitive processes that students need to develop?

• Might AI tools provide shortcuts that could hinder the development of fundamental skills or understanding?

• Are there aspects of the learning process that are crucial for students to experience without AI assistance?

There may be critical thinking or problem-solving skills that could be short-circuited by AI assistance. It's important to assess how AI use might affect students' core abilities to engage in deep thinking about your field’s concepts, to solve problems, and to interact with colleagues, including those abilities “down the road” in future courses or careers.

Next, examine the potential impact on fundamental skills and understanding. AI tools might provide shortcuts that could hinder the development of basic competencies in your field. Consider whether there are foundational skills that students need to master without AI assistance. It's also worth pondering if reliance on AI could obscure gaps in students' essential knowledge.

Creativity and original thinking should also be taken into account. Consider whether AI tools might inadvertently limit students' creative problem-solving or innovative thinking. Reflect on how you can ensure that AI enhances rather than replaces students' original ideas.

Long-term skill retention is another important consideration. Contemplate whether AI dependence could lead to a decay of skills or knowledge over time. Think about how AI use today might impact students' ability to perform tasks independently in the future.

Also, reflect on the essential learning experiences in your course. There may be aspects of the learning process that are crucial for students to experience without AI assistance. Think about hands-on or experiential learning components that should remain AI-free. Consider how AI use might impact students' ability to engage with primary sources or raw data in your field.

➕ Step 3: Consider Adding New Objectives

Even if your existing course objectives were comprehensive for the pre-AI era, there may be ways that they need to be supplemented given the advent of AI.

For instance, if you adjust an existing course objective to account for the significant role that an AI tool will play in achieving it, you may need to add further objectives for training students on the fundamentals of using this AI tool.

It is likely these fundamentals will require theoretical understanding and practical proficiency.

You may also need to add new objectives about the ethical use of AI — that have no analogues in your prior objectives — because AI introduces a range of new moral considerations that weren’t relevant previously.

In short, there are three main sources of new course objectives related to AI use itself:

1. those that concern your students’ theoretical understanding of AI

2. those that concern their practical proficiency with AI

3. those that concern their ethical use of it

When developing learning objectives on these three fronts, I highly encourage you to be careful to dip down only to the layer of the AI waters as you need for your course. Your goal should be to ensure students are not overwhelmed with unnecessary technical details.

In other words, as you consider adding AI-specific objectives, it's important to remember that the goal is not to turn every student into an AI expert or data scientist (unless, of course, your field is AI itself or data science). Rather, the aim is to equip students with the knowledge and skills they need to effectively and responsibly use AI tools within your specific discipline, and to critically evaluate the outputs and implications of these tools.

Indeed, in one of my early attempts to train my students to use AI, I can report that many of my students told me later that they felt overwhelmed when I provided too much technical detail on how LLMs work — and this detail wasn’t really relevant to the uses of AI that we focused on. This was a learning moment for me.

Consider Adding New Objectives for
Theoretical Understanding of AI/LLMs

Since I am not an expert in fields other than my own, I will focus this subsection on the sort of foundational theoretical understanding that is essential for all students using AI, regardless of their field of study.

You should add or adjust the following to your specific needs.

At the most basic level, students need at least a rough working definition of AI, an abstract understanding of its most relevant form in the current paradigm (namely, machine learning), and some common applications of this form relevant to the field of study or course.

So, I recommend that you consider including new objectives that require your students to be able to…

Subscribe to Premium to read the rest.