Personally, I’m a fan of good comprehensible input and a fan of well-done lessons.
I’ve been using LingQ several years yet I’ve done several hundred structured lessons with Lingoda and Babbel Live.
While I’m working on a general-purpose ChatGPT prompt to take any language lesson PDF and turn it into different kinds of comprehensible input that relate to the themes, vocabulary, grammar, and more of a lesson, here’s an example of a hand-written prompt and simple follow-up for a given lesson on the Internet and social media.
Take this French language lesson and take its themes, ideas, and vocabulary and generate 2,500 words of content in French at the B2 level. Create a panel discussion where the panelists talk about the use and effects of social media and the Internet in general. Have the discuss the challenges and opportunities of the generations–children and gaining awareness of the world and the use of technology at school, young adults and the implications of technology for career choices and in finding romantic relationships, parents and their role in helping both their children and older generations navigate contemporary tech, and the eldery and their unique opportunties and challenges related to vulnerabilities and scams.
Then after ChatGPT 4o’s first pass…
How many total words did you generate?
It generated only around 1,000 words when I asked for 2,500. Didn’t surprise me, ChatGPT likes to work in smaller chunks. I followed with.
Expand to reach the target length. Search the internet for actual research, surveys, and other facts you can have the panelists cite as experts. Further, search the internet for anecdotal stories that may heighten the emotional aspects as well. Use this approach to add more content.
Then I followed with…
Thank you. Provide a link to download as a TXT file.
And then…
And finally, please suggest a 60-character or less title, brief subtitle/description and generate a simple, landscape dimension image that illustrates the text.
I import into LingQ. While ChatGPT didn’t get to the desired word count in two passes, in this example, after generating the audio, it’s about 7 minutes of audio, 9 new words, 7 LingQ, and 425 known words.
Over recent months, I’ve been spending about an hour to prep for each class. I’ve prepped, in many ways, like a traditional student where I do the exercises in advance, look up unknown vocabulary, answer closed questions and jot down notes to answer open questions, etc. This does make me look like a model student and gives me positive feedback for that, but I think I’m going to experiment with using AI-generated comprehensible input to prepare instead.
While I might have to think more in class, I think I might be able to engage with better fluidity too.
I’ve spent a lot of time over recent weeks with ChatGPT in both language learning and in my job. I’m getting to learn how to engage it better to produce better and better results, but I’m growing more and more frustrated with it in ways that one can get frustrated with a weak team member on a group project. It can be a lot of work to get them to do good work.
I wonder, can traditional lessons and comprehensible input be synthesized yielding synergies.