blogger counters

Sunday, May 05, 2024

How to Effectively Design Good Generative AI-Using Student Assignments: Steps, Strategy, and an Example

If it were not bad enough to combat essays purchased from Course Hero, GradeSaver, or one of the other paper mills that purchase student essays, now one has to worry about papers generated by large language models such as ChatGPT, Scribe, Google Gemini, CoPilot or others.  Generative AI tools are being developed for specific topics or domains, and the assumption is, at least from a student’s perspective, that the product will be well written and accurate, or at least well written and accurate enough to merit a passing grade. 

As an instructor who develops, teaches, and grades courses and coursework for online courses hosted on Canvas, Moodle, Blackboard, or even Google Classroom, this is a challenge you can’t afford to ignore, and an opportunity that can transform your entire teaching philosophy. 


What data sets are used to train the generative AI algorithms?  Let’s take the case of American literature. I would automatically assume that the first repositories to be ingested would be open access repositories such as Project Gutenberg and  Those would be primary texts for the most part. Then, secondary texts would be incorporated, which would consist of journal articles. Tertiary texts would also be used, which would include encyclopedias. While there may be some question about intellectual property and the right to use the materials, there would be few questions about the integrity of such materials, gleaned as it were, from peer-reviewed and quality assured primary sources, peer-reviewed journals and monographs, and peer-reviewed encyclopedias. 

However, this is just the tip of the iceberg when it comes to materials about American literature that one can access through the Internet. There are blog posts, online magazines, repositories of articles from study guide providers such as, Sophia Learning,, and many more. These resources may be a bit repetitive and not particularly original, but perhaps harmless in the overall scheme of things.  If anything, they may skew the results to bland and formulaic papers. 


What happens, however, when these AI bots ingest the massive repositories of essays which have been compiled by companies such as Course Hero, Grade Saver, and others?  Is there any quality control at all? Some of the sites encourage students to sell their term papers for $5 each. Is there any quality control?  Would the evaluator or screener check for accuracy, bias, originality, and innovative thinking? Would the screener be able to differentiate between a tired, flat, biased, inaccurate paper and a fresh, innovative, and accurate one? 

It is quite possible that the sheer volume of low-quality term papers, biased or error-riddled blog posts, student-shared material, unrefereed conference proceedings and “gray” literature could exceed the high-quality, peer-reviewed material by many orders of magnitude. 


Learning how to use generative AI as a helpful tool and understanding its limitations and pitfalls is probably a more pragmatic and useful approach than trying to ban it altogether. Further, if students learn how to use AI tools in their coursework where they receive feedback and guidance, it is likely that they will be able to more effectively use the tool in their professional lives. 

Instructional designers and instructors who understand how results are likely to be generated can design assignments that require students to incorporate their own unique vantage points, prior knowledge, experience, and insights rather than simply producing a bland summary or compendium of the blandest, least thought-provoking or original material that responds to the prompt. 


Instructors can develop assignments that actually require students to use a generative AI tool, and then they can ask them to critique the response.  

Here’s a sample assignment: 

Please write an essay that identifies the possible themes in Emerson’s essay, “Circles,” and explains how they relate to American Transcendentalism. 

When I fed the prompt to ChatGPT, it churned out a six-paragraph essay in about 3 seconds. It sported a nice, clear thesis statement, a clear definition of American Transcendentalism, and then proposed several themes, and provided evidence, including quotes. On first glance, it seemed to be a very serviceable essay. 

However, upon closer examination, it was clear that the essay was pretty facile, and felt very derivative. It did not provide citations for the quotes, and the explanations around each of the quotes and the themes did not go into any depth and tended to repeat each other. There was no deeper probing of American Transcendentalism, nor was there any historical context. 

One could argue that the prompt should have asked for historical context, and could even be interesting and ask for connections to a darker, more ominous underbelly of what seemed, on the face of it, to be an expansive, all-encompassing dream of harmony and unity. However, there is no investigation of the philosophical or ideological underpinnings, nor any sense of how American Transcendentalism was popular because it uniquely reinforced and even expanded the authority of those who already possessed it.  

The students could be asked to identify the weaknesses of the Chat GPT essay, and to explain when and how the responses were cliché, facile, or derivative of what one would expect to see in an encyclopedia or generic study guide, and they do not encourage critical thinking or a look at “Circles” through a new and innovative vantage point. 

It would be very nice for students to work on the critique in a collaborative workspace – for example, in a shared document, where each could highlight and comment on the ChatGPT-generated essay. 


It is possible to write assignments and paper prompts that encourage students to incorporate their unique views, experiences, and prior knowledge.  While some may think that this subjectivity would make it difficult to assess the work, in reality, just the opposite is the case. 

For example, for an essay on American literature, a rubric would require the student to accurately describe what the text is about and also the main themes and characters as required by the prompt.  The prompt should also, however, require the student to think critically and ask questions that would relate to their own experiences and perceptions. They could also be asked to apply the concepts to a current situation. The students would be evaluated on their creativity and problem-solving skills as well as critical thinking. 

To apply this to the case of Ralph Waldo Emerson’s essay, “Circles,” a prompt could look something like this: 

Evaluate the strengths and weaknesses of Emerson’s ideas about agency as expressed in his essay, “Circles,” and describe how you’ve seen examples of the thinking expressed in your own life or the world at large.  Describe mid 19th-century American Transcendentalism and how it must have sounded to different groups of Americans: wealthy land owners, men wanting to move West for wealth and adventure, women in America, Native Americans, African-American slaves and freedmen, Irish indentured servants, Mexicans living in parts of Texas that were a part of Mexico before the War of 1848, and children. 

Granted, the new prompt is much, much more complex, but you probably get the picture, and can see how writing prompts to outwit the  AI bot is fun and engaging, while also encouraging collaborative activities and deeper learning. 


Colleges and universities are grappling with ChatGPT, Scribe, Google Gemini, CoPilot, and other generative AI platforms and tools.  While there is the chance of rampant violations of academic integrity, we are in the early stages of development and now is the time to envision new and productive pedagogical approaches that are ideal for adult learners and can help develop creativity, critical thinking skills and confidence.  

Blog Archive