When AI is not the (best) answer

International Academic Integrity Day, marked annually in October, is about honesty, trust and fairness in academic work. 

With the spread of artificial intelligence, including generative AIs like ChatGPT, it’s important for students to use AI thoughtfully and only as allowed within your UM course guidelines. 

As you're studying, collaborating with peers and completing coursework, here are some things to consider. 

Double-check what’s allowed

Different instructors and courses may have different approaches to the use of AI. Some may prohibit AI use in their course. Others restrict use of AI to parts of certain assignments. Always check what is allowed in your course and on specific assignments.

Each assignment, course and instructor will have different guidelines for AI use, depending on your program of study; there is no one-size-fits-all rule. 

Using research methods vs. AI

Though AI is very powerful in gathering and synthesizing information, this is also its limitation.

AI pulls from the entire internet. It draws on a wide swath of information of varied quality and from many sources, including potentially dubious, inaccurate or irrelevant sources. Research follows a specific methodology or specific, contextual processes.

Verifiable facts and information — credible information that can be shown to come from credible sources — are central to academic writing and research.

AI can be wrong

ChatGPT sounds very confident when it spits out that answer to your question about chemical interactions or concocts a quick essay on the significance of a historical event.

But AI can hallucinate or be off the mark in divergent and significant ways. AI-generated content can contain factual errors; it can be inaccurate or incomplete. AI can misinterpret texts or cite sources incorrectly, hallucinate or entirely manufacture information.

It’s important to understand how to evaluate information and to be able to reflect critically on the quality of information, including information generated by AI.

Tiana Thomson, an academic integrity coordinator advising UM students, sometimes offers this quick test for AI use: “Do you have the knowledge base to evaluate a GenAI output before using it in your work?” she asks.

“If you don't have that knowledge base and don't plan to spend the time evaluating the information for accuracy, reliability, and relevancy, then AI is probably not the best tool for the task!”

AI writing sounds ‘smart’ but fake

Another reason to be wary is because AI-generated responses often sound generic and overproduced. AI writing relies on repetitive structures and cliches and often uses predictable or obvious points.

This can make AI-produced work sound “smart,” but feel empty or oddly smooth and impenetrable. It lacks the individual thought processes, creative expression and contextual connections that make something your own unique product.