Collaborative Fact Checking of AI-generated Text

With all the news about ChatGPT and the new AI tools around I’m mostly interested in how to integrate them into my teaching. As usual, very good ideas come from David Wiley who recently stressed the opportunities for instructional designers in this new era of AI.

The key point with all the stuff that comes out of ChatGPT and the tools that will follow seems to be fact checking. In this delta lies an opportunity for learning if you don’t take the forensic approach trying to find evidence for the AI lying. I suggest taking the generated text as a baseline to start from.

Chaining tools for fact checking

Hypothesis and HedgeDoc are a good combination for fact checking generated text by AI. I just have the idea of the following assignment:

1. Generate a text with ChatGPT
2. Open a new HedgeDoc.
3. Paste <style>.markdown-body {margin-left: 0}</style> at the very beginning of the pad in order to give room to the Hypothesis panel on the right hand side.
4. Paste the generated text in the pad.
5. Switch to the reader view in HedgeDoc.
6. Paste the link URL of the HedgeDoc in Hypothesis.
7. Start fact checking, annotating and discussing collaboratively in class.

I’m looking forward to trying this out.