Interacting with Generative AI - Prompting Effectively

At the end of this section, you should be able to distinguish between an AI prompt and a traditional keyword or Google search to choose the most effective tool for academic tasks.

At first glance, prompting an AI tool and typing a query into Google might feel similar. But they operate in fundamentally different ways.

Here are some key differences to consider:

Choosing the right tool for your academic task can significantly impact the quality, reliability, and credibility of your work. Search engines like Google or Google Scholar are best suited for finding accurate, up-to-date, and verifiable sources, especially when you need to cite evidence or support claims. Generative AI tools, on the other hand, are helpful for brainstorming, summarizing complex ideas, or exploring unfamiliar topics. However, their responses can sometimes be inaccurate, outdated, or difficult to verify, so fact-checking is essential.

By understanding the strengths and limitations of each tool, you can make informed decisions, work more efficiently, and meet academic expectations for both accuracy and integrity.

While earlier AI models (like ChatGPT-3) generated responses solely from pre-existing training data, newer tools, like Perplexity and Copilot, use a method called retrieval-augmented generation (RAG) to do both, generate and retrieve. Hybrid tools combine elements of generative models (that produce responses based on training data) with retrieval-based models (that access real-time, external information). This approach can improve accuracy, reduce hallucinations, and provide more up-to-date information.

However, even is a tool uses RAG, outputs still require critical evaluation. The quality of retrieved sources can vary, and AI may misinterpret or misrepresent the information it finds. As with any academic tool, your judgment is essential. Always verify claims, cross-check with original sources, and consider whether the information meets the requirements of your assignment or the standards of your discipline.

Before selecting a tool for academic work, ask yourself:

While many generative AI tools are available online, not all of them prioritize your privacy or align with university policies. When using AI tools for academic purposes, especially for learning, research, or group projects, it’s important to choose a tool that is vetted and safe.

⚠️ Avoid using public AI tools for academic tasks unless you fully understand their privacy policies and terms of use. Public tools may collect your data or use your input to train future models, which can raise serious privacy and intellectual property concerns.

Accessing Copilot Follow these instructions in PAWS to access the USask version of MS Copilot. Look for the green shield once you have logged in.

🛡️ Tip: To know if an AI tool is RAG enabled or searching the web in real-time, look for signs like citations or numbered references at the end of a response, source links you can click to view the original website, or mentions of very recent events (e.g., news from the last few days). These clues suggest the tool is retrieving up-to-date information from the web. However, even when using a USask-approved tool like Copilot, it’s still your responsibility to think critically and verify the information.