Experiment: I created an entire lesson package on cognitive dissonance with AI
It includes an explainer, a slide deck, activity ideas, MCQs, take-home assignment options, and a FAQ handout.
As you might have noticed by now, I have been playing with AI to explore how this technology can help teachers and learners in our field.
Yes, it’s not all roses. I can foresee AI cheating will be a challenging and prevalent issue, given how fast the tech is improving (this article says ChatCPT went from a D grade to an A grade student in an economics exam in just three months, for instance).
It is a powerful, undetectable plagiarism tool, too. To prove a point, some researchers published a paper about AI that they had ChatGPT write! Generative AIs also frequently ‘hallucinate’ (make stuff up) far more often than they should, including some bogus Guardian news articles.
Nonetheless, many journalists and educators have been trying to take advantage of its potential. For example, I found this beginner’s handbook written for small news publishers very useful.
Generative AI in education is a hot topic, and many academics focusing on pedagogy in different fields have written about its future implications and good use cases. I am also convinced that this technology will drastically change education in the near future.
Creating a full lesson package
So, after coming back from almost a week-long Ching Ming - Easter holiday in Osaka (and gaining a few pounds), I tasked myself with creating a comprehensive one-hour lesson plan to teach cognitive dissonance in a journalism class using AI.
In this newsletter, I will discuss the steps I took and how I used ChatGPT to develop this package containing:
an explainer,
a 15-page slide deck,
class activity ideas,
MCQs,
and take-home assignment options.
Step 1: Establish a premise to generate a ‘prompt’
This first step is crucial. It is a well-documented fact that ChatGPT and other LLMs (large language models) produce quite different results depending on how you form your questions and instructions.
‘Prompt engineering’ is reportedly a highly sought-after skill at the moment for this reason. According to a Bloomberg article, an ‘AI whisperer’ job could pay 335,000 USD.
But, of course, I am by no means an expert. What I did first was to establish the premise as follows:
Objectives: Learners understand 1) what cognitive dissonance is, 2) how it relates to news consumption as well as news production, and 3) what to do about it.
Target learners: University students in a fact-checking class who may have little or no knowledge of psychology concepts.
Methods: A brief primer (lecture) including a definition of cognitive dissonance with news examples, followed by discussions, a mini-test (MCQs) and a take-home assignment.
Step 2: Develop an explainer with the prompt
Based on the above, I’ve used the following prompt to generate an explainer that discusses cognitive dissonance.
[Prompt] As an expert in cognitive psychology, your task is to explain the concept of cognitive dissonance to a university lecture audience consisting of 17-21-year-olds. Your explanation should focus on how this phenomenon affects news production, dissemination, and consumption as part of a course about fact-checking. Please provide a clear and concise definition of cognitive dissonance that includes relevant examples and research findings. You should also describe how cognitive dissonance can impact people's beliefs, attitudes, and behaviors related to news production, consumption and sharing. Additionally, you should discuss strategies for identifying and reducing cognitive dissonance in oneself and others when it comes to news. Your response should be engaging and accessible to students with varying levels of familiarity with psychology concepts. Please note that you may use multimedia tools or interactive activities to enhance your presentation if desired.
I know I know, it looks rather convoluted, but this is how you should ‘speak to’ generative AI to get better results.
If you’re new to this, this website called PromptPerfect could be extremely helpful.
It generates prompts like the above from a much shorter, one- or two-sentence instruction and lets you compare the results.
I honestly think “there is an AI for that” will be a new meme line. Anyways, here’s the link to the Explainer it generated.
Step 3: Create a slide deck
To complement the explainer, I wanted to create a slide deck consisting of 12 to 15 pages and used the following prompt:
[Prompt] Make a set of 12 to 15 slides based on the explanation above. Each slide only conveys one idea in a clear and concise manner. Give ideas about possible charts, illustrations and other visual aids when appropriate.
ChatGPT then created a 15-page slide deck idea, which is not perfect, as you can see, but I believe any educator will see the value of having this as a starting point. At a minimum, you can add some local news examples to it, and the deck is ready to go.
If you like, there are many AI-driven presentation makers that generate nice slide templates for you as well. Beautiful AI, Canva, and a number of other services are available or soon-to-be-available, seemingly including PowerPoint and Google Slides.
I am not sure what design fits such a topic as “cognitive dissonance and journalism,” but those AIs will figure that out and generate a series of different look-and-feel for you to choose from.
Some of them even fill the slides with suggested information about the topic without your giving any instruction.
Step 4: Generate activity ideas
The next step is to develop possible in-class activities. I instructed the AI as follows:
[Prompt] Based on the above definition of cognitive dissonance, create three learning activities that can enhance the learners' understanding of the concept and how it relates to fact-checking.
ChatGPT gave me a variety of ideas — a role-play, a fact-checking exercise, and a discussion — but they are all based on the assumption that the students are simply the news audience.
Many students I have are also active disseminators and producers of the news as well. So, I further instructed:
[Prompt] These examples and activities are all for the news audience. Generate three more activities to focus on the production and dissemination of the news.
It then suggested a newsroom simulation, news source analysis, and even a panel discussion with professional journalists.
Again, I believe these ideas give a good starting point for teachers to think about the best activities for their learners to relate their personal experiences to the concept of cognitive dissonance and think about it in relation to news, journalism, and fact-checking.
Step 5: Produce MCQs
Testing students with easy multiple-choice quizzes is one way to recap the class and evaluate how effective the lesson has been.
In this experiment, I asked ChatGPT to generate ten questions:
[Prompt] Make 10 MCQs based on the idea of cognitive dissonance defined in this chat previously with explanations.
Here is the result. These MCQs included explanations of the correct answers, as well as some additional information. Because we are dealing with just one topic, the questions are somewhat repetitive, but it is not necessarily a bad thing. It may help solidify learners’ understanding, in fact.
But here, I realised we also need to look at each question, answer, and explanation very carefully before using the material because some questions can possibly confuse the learners as well.
For example, the difference between “a journalist selectively reporting information that supports their existing beliefs” and “a journalist ignoring conflicting information that challenges their existing beliefs” is very nuanced (question number 8) and perhaps an unnecessary distinction in our context.
A few other questions may not be usable as is and require rewording before giving them to the students.
Step 6: List take-home assignment options
Finally, I had the AI develop three take-home assignment ideas.
[Prompt] Give three ideas for take-home assignments that will deepen the understanding of the concept in relation to the recent news stories and social issues.
One of them was a reflective journal, which isn’t particularly original, but I got curious and asked ChatGPT to write a sample essay.
[Prompt] Write a reflective journal example in 300 to 500 words including a brief summary of what the news stories are about and what information triggers the cognitive dissonance and why. Avoid examples related to news stories about public health and politics.
I excluded health and politics as a news example topic because if I were teaching this imaginary class for real, the chances are that I would perhaps mention news related to the pandemic and local politics in my lecture.
ChatGPT generated an essay as a student discussing how a news story about a company polluting the environment triggers cognitive dissonance because she/he likes the products the company produces and buys them often.
Not a bad example and quite realistic.
If a student submits something like this with real news examples, I would give him/her a pass (after all, it’s just homework), although the last two paragraphs sound very pretentious.
Step 7: Make a handout with FAQs
This last step is just the icing on the cake. I knew it would just generate the same stuff included in the explainer and MCQs, but I tried anyway because creating FAQs is one of the things generative AIs are good at.
[Prompt] Create one-page FAQs to distribute to the students after the class.
As predicted, ChatGPT gave me a nice set of six questions that fits nicely on one page.
My thought
Cognitive dissonance is something I have been teaching for a while now. I have my own sets of slides, activities, MCQs, and other teaching and learning materials surrounding this idea.
Compared with those, which are full of real news examples and much more focused on journalism and fact-checking, what the AI produced is relatively generic. It is not good enough for my class, but I can also see this set might be really handy for some other educators whose objectives are different from mine.
I should note that, unlike some other AI tools like Bing Chat and Perplexity, the version of ChatGPT I used (the free one) does not give links to references (rather, it makes up citations when asked), so I did not bother to generate a reading list.
To be honest, it took some time for me to tweak the language here and there and try out different prompts. It wasn’t much of a time saver. I could have made something on my own in the same period of time without the help of AI.
But I am starting to get the knack for it. I am sure it will be faster the next time. This experiment has somewhat solidified my conviction that the role of artificial intelligence in education will become exponentially important in the near future.