One of my favorite things about shifting to a remote or hybrid learning setup are discussion posts. I know that not every student is fond of them, but I’ve always been a big fan of healthy discourse between classmates.
However, I don’t always have the time to respond. More than that, I find that my enthusiasm is often low when a professor requires a reply as an assignment. Suddenly, discourse becomes a tedious task.
When that happens, I rely on LLMs to write my responses. But here’s the kicker: I don’t use it to cheat. Instead, I use it as a writing assistant.
Want to learn how? In this article, I’ll be going over how to create and humanize discussion post responses for Blackboard, Canvas, Google Classroom, and more. Stay tuned!
Step #1: Make An Outline of Your Answer
When asking ChatGPT or other LLMs to provide a discussion post answer, it’s best to already have an outline in mind. This should contain the following:
- The topic.
- Main talking points.
- Supporting information.
- Sources.
The reason for this is simple: LLMs have a tendency to hallucinate. You can’t expect them to generate real data with 100% accuracy. That’s why you need to assist them as much as you can with the information and let them deal with the final output.
If you need an LLM that can generate information using real-time data, it’s best to use Gemini, but always double-check their responses.
Step #2: Generate Your Content
Once you have your outline, it’s time to build your prompt. I’ve tried many prompts, and this is the one that works best for me:
Hi, [LLM Model]. Pretend you’re a student and you’re replying to a discussion post. Craft a good response. Write in first-person. These are all the information you need: Topic: [Topic] Question Posed: [Question] Supporting Information: [Any helpful information along with sources] Make sure you’re writing as a student and not as an LLM. Add a level of uncertainty. Pose questions to encourage discourse. | Hi, ChatGPT. Pretend you’re a student and you’re replying to a discussion post. Craft a good response. Write in first-person. These are all the information you need: Topic: The effects of global warming to the global south. Question Posed: Do global south countries disproportionately get affected by rising temperatures? Supporting Information: – The Global Climate Risk Index consistently ranks many countries in the Global South as most susceptible to climate damages, despite low emission levels [source: Global Center for Ecology and Development]. – Up to 75% of the food supply in developing countries comes from smallholder farmers, who are most vulnerable to disruptions caused by extreme weather events [source: Global Center for Ecology and Development]. Make sure you’re writing as a student and not as an LLM. Add a level of uncertainty. Pose questions to encourage discourse. |
Here’s what the output looks like when I use the sample prompt on ChatGPT:
Hi everyone, I agree that global south countries are disproportionately affected by rising temperatures. The World Bank emphasizes that the poorest and most marginalized suffer the most from climate change. Extreme weather events, food and water insecurity, displacement, and cultural threats are particularly severe for these communities. They often lack the financial resources, infrastructure, and technology to adapt effectively. Furthermore, the Global Climate Risk Index consistently ranks many Global South countries as highly susceptible to climate damages, even though they have low emission levels. Considering that up to 75% of their food supply comes from smallholder farmers, who are highly vulnerable to extreme weather disruptions, it’s clear that these nations face significant challenges. However, I’m curious about the potential solutions. How can international support be better directed to help these countries? Are there successful case studies of communities in the Global South effectively adapting to climate change that we can learn from? Looking forward to hearing your thoughts! |
Looks great, right? You could send this already, but some educational platforms prevent AI use. So, that’s why you need to humanize these further.
Step #3: Use Undetectable AI
Undetectable AI is a popular AI bypasser that changes AI-generated text into something that can trick detection tools. Using this with your discussion posts changes it so it’ll appear as if a real person wrote it.
One of the many features of Undetectable AI is output customization. Always keep this in mind when you’re using it for discussion posts. I’ve played around with it a bit, and I find that using “University” for readability, “Balanced” for balance level, and “General Writing” for purpose works best in this use case.
Here’s what our output looks like when it’s processed with Undetectable AI:
Hello everyone I share the belief that countries in the global south bear an impact from increasing temperatures. The World Bank highlights that those who are poorest and most marginalized endure the effects of climate change. These communities face challenges such as extreme weather events, food and water insecurity, displacement and threats to their cultures. They often lack the means, infrastructure and technology needed to adapt effectively. Moreover the Global Climate Risk Index consistently identifies nations in the Global South as highly vulnerable to climate related damages despite having low emissions. Given that around 75% of their food comes from small scale farmers who’re particularly at risk from extreme weather disturbances it is evident that these countries encounter significant obstacles. I am intrigued by solutions. How can international aid be better channeled to support these nations? Are there examples of communities, in the Global South effectively adjusting to climate change that we could draw lessons from? Looking forward to hearing your perspectives! |
A quick look will tell you that this output isn’t perfect. This is because Undetectable AI automatically puts mistakes to make the text appear more human. We’ll fix this in a bit, but before we move on, you can read our full review of this platform here and some of our other articles like this one.
Step #4: Tweak The Output Further
The last step is to tweak Undetectable AI’s output. This includes removing the intentional errors and contextualizing the output based on how people know you in class. You should also remove any common AI identifiers that are still present in the text, like:
- Common AI words. This includes words like “delve”, “galvanize”, “utilize”, and transition words.
- Repetition. Multi-paragraph AI texts usually have repetition across different paragraphs.
- Lists. Most LLMs have an issue where they add more information through lists without expounding.
Here’s what the final output looks like once I tweaked it:
Hello everyone, I believe that countries in the global south shoulders more impact from increasing temperatures. The World Bank highlights that those who are poorest and most marginalized endure the effects of climate change more. These communities face challenges such as extreme weather events, social insecurity, displacement, and threats to their cultures. They often lack the means to adapt effectively. The Global Climate Risk Index also consistently identifies nations in the Global South as highly vulnerable to climate related damages despite having lower emissions. Given that around 75% of their food comes from small-scale farmers who are particularly at risk from extreme weather disturbances, it’s evident that these countries encounter significant obstacles. As such, I am intrigued by solutions. How can international aid be better channeled to support these nations? Are there examples of communities in the Global South effectively adjusting to climate change that we could draw lessons from? Looking forward to hearing your perspectives! |
Step #5: Press Post
And that’s about it. All you need to do now is post it and wait for your classmates to provide their own input.
Should You Ask ChatGPT to Make Discussion Post Responses?
AI’s role in education has always been murky at best. Students’ overreliance on ChatGPT, in particular, has been criticized due to its lack of governance (unlike Claude) and constant hallucinations. But I don’t believe it’s fair to ask if you should use ChatGPT to make discussion posts responses — it’s all a matter of framing the question.
Instead, you should ask how you should use ChatGPT to make discussion post responses.
GPT-4, Gemini, Claude Sonnet and Opus — these are all amazing study companions if you use them properly. Let them create outlines, generate ideas, or host quizzes. You can be a better student by using them for anything except simple copy-pasting.
All Said And Done
As always, practice responsible AI use.
LLMs should never be used for educational purposes alone. It’s best to integrate them into your workflow and make your studying habits more efficient. Remember that it’s not the teachers that suffer the most from using ChatGPT to cheat, but students who are left behind due to overreliance on AI models.
Discussion posts are an avenue and a safe space to share your thoughts — so share them freely.
Interested in learning more about responsible AI use in education? May I suggest articles such as this one? Happy learning!
