How Text-Generating Software Still Needs Human Guidance
ChatGPT and other text-generating software have undoubtedly appeared on your radar in some form to support your teaching practice. Perhaps you’ve heard about how quickly it can help you create a rubric for your course or how it can help you brainstorm topics for classroom discussion. Text-generating software may speed up the course creation process. Still, as my colleague Eric Freeman and I recently discussed in a presentation, we must first slow down to avoid any pedagogical complications. In this brief post, I’d like to share some significant takeaways from our presentation.
Key takeaway #1: Your output in only as good as your input!
Be intentional about what you want the text-generating software to produce. Be specific about your desired output by defining clear parameters around things like subject matter, audience, quantity of questions, length of time, etc. If the prompts you enter are too general, you will receive content from the text-generating software that is equally as general. As a result, the material may be irrelevant or too broad to apply to your work.
Example: I want ChatGPT to write three multiple-choice questions about Canadian women's involvement in the Second World War with feedback for learners. In this case, I've thought carefully about the type of question I want (multiple choice), how many questions I want (three), what the subject is (Canadian women's involvement in the Second World War), and that I want feedback for each question. Very quickly, the program produces results. I may want to check these questions for accuracy and perhaps tailor them based on the content of my lesson and audience.
Key Takeaway #2: Text-generating software can save us time - so long as we carefully review and alter outputs for quality
Text-generating software can do its best to produce material for you based on the prompts you input, but just because it outputs content doesn’t mean that content is accurate or entirely useful for your purposes. Regardless of your expertise on the subject, you mustn’t take any AI-produced content at face value and you should check it for accuracy and relevancy.
Key Takeaway #3: AI is not a pedagogical expert
Think strategically about the inputs you insert into text-generating software to ensure they reflect instructional goals and outcomes. Well-constructed inputs can support you in simplifying the course design process. However, AI doesn't have pedagogical expertise. It is on the creator to:
- ensure learning outcomes are relevant, actionable, and measurable (if asking AI to design them).
- review content and ensure it is accurate, related to learning outcomes, and chunked appropriately to reduce the cognitive load.
- check that any assessments created by text-generating software allow learners to demonstrate mastery of learning outcomes.
How have you used text-generating software to support your course creation? In what ways has it been helpful? What strategies do you use to ensure any material produced by text-generating software is pedagogically sound?
Sources:
Image Credit:
- Photo by Matheus Bertelli from Pexels: https://www.pexels.com/photo/a-man-is-typing-on-a-laptop-computer-16094048/